PROGRAM OF
The 118th Meeting of the AcousticalSociety of America
Adam's Mark Hotel ß St. Louis, Missouri ß 27 November-1 December 1989
MONDAY EVENING, 27 NOVEMBER 1989 ST. LOUIS BALLROOM D, 7:00 TO 9:00 P.M.
Tutorial on Architectural Acoustics
Mauro Pierucci, Chairman Departmentof Aerospaceand EngineeringMechanics, San DiegoState University,San Diego, California 92182
TUI. Architecturalacoustics: The forgottendimension. Ewart A. Wetherill (Wilson, lhrig, and Associates, lnc., 5776 Broadway,Oakland, CA 94618)
The basic considerationsof architeclural acouslics---isolation from unwanled noise and vibration, control of mechanicalsystem noise, and room acousticsdesign---are all clearly exemplifiedin Sabinc'sdesign for BostonSymphony Hall. Openedin ! 900,this hall isone of theoutstanding successes in musical acoustics. Yet, aswe approachthe hundredthanniversary of Sabine'sfirst experiments, acoustical characteristics remain one of the leastconsidered aspects of buildingdesign. This is due, in part, to the difficultyof visualizingthe acouslicaloutcome of designdecisions, complicated by individualjudgment as to whatconstitutes good acous- tics. However,the lack of a comprehensiveteaching program remains the dominantproblem. Significant advancesover the past 2 or 3 decadesin measurementand evaluationhave refinedthe ability to design predictabilityand to demonsIrateacoustical concerns to others. New techniquessuch as sound intensity measurements,new descriptors for roomacoustics phenomena, and the refinemen t of recording,analysis, and amplificationtechniques provide fresh insights into the behaviorof soundin air and other media.These topics are reviewedwith particularemphasis on the needfor a comparableadvance in translationof acousticprinci- plesinto buildingtechnologies.
Sl J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of America S1 TUESDAY MORNING, 28 NOVEMBER 1989 ST. LOUIS BALLROOM C, 8:00 A.M. TO 12:00 NOON
SessionA. Architectural Acoustics I: Electronic Room Simulation for Production and Reproduction
Elizabeth A. Cohen, Cochairman CharlesM. Salter Associates,Inc., 130 Sutter Street,San Francisco,California 94101
J. ChristopherJaffe, Cochairman JaffeAcoustics, Inc., 114A WashingtonStreet, Norwalk, Connecticut 06854
Chairman's Introduction4:00
SpecialTribute--8:05 Theodore John Schultz--A Tribute Presentedby:Ewart A. Wetherill ( Wilson, lhrig, and Associates, 5776 Broadway, Oakland, California 94618)
Invited Papers
8:15
A1. Electroacousticalsimulation of listeningroom acousticsfor projectARCHIMEDES. S6rcn Bcch (The AcousticsLaboratory, Technical University of Denmark,Building 352, DK-2800 Lyngby,Denmark) ARCHIMEDES is a psychoacousticsresearch project, funded under the EuropeanEUREKA scheme. Threepartners share the workinvolved: The AcousticsLaboratory of The TechnicalUniversity of Denmark; Bangand Olufsen of Denmark;and KEF Electronicsof England.Its primaryobject is to quantifythe influence of listeningroom acoustics on the timbreof reproducedsound. For simulationof theacoustics of a standard listeningroom, an electroacousticsetup has been built in an anechoicchamber. The setupis basedon a computermodel of thelistening room, and it consistsofa nmnberof loudspeakerspositioned on an imaginary spheresurrounding the positionof the testsubject. The setuphas been designed for the highestdegree of flexibility.This includesthe possibilityof simulationof directivitycharacteristics of normaldomestic loud- speakersand absorption coefficients of thesurfaces of thelistening room. This paper is a presentationof the system,with special emphasis on the psychoacoustical background of thedesign. This will includea discussion of choiceof experimentalprocedure, test stimuli, and test subjects as well as purpose built loudspeakers and the DSP system.
8:45
A2. The psychoaeousticsof loudspeakersound reproduction--Past achievements and present problems. Floyd E. Toole (Division of Physics,National ResearchCouncil, Ottawa, Ontario K1A 0R6, Canada) The designof loudspeakersis graduallybeing put on a scientificbasis. Art and intuitionhave given way to engineeringguidelines as the relationshipsbetween perceptions and technicalmeasurements have been eluci- dated.Indeed, within limitedcircumstances, loudspeakers can be designedto meetspecific engineering design objectives,with considerable confidence in howthey will besubjectively evaluated. In practice,however, all of therelevant conditions are not controlled, and several factors conspire to precludeuniversal satisfaction among listeners.Most of the uncertaintyappears to bein the interactionsbetween loudspeakers, rooms, listeners, and programmaterial. This paperreviews the presentstate of knowledge,and outlinesthe areasmost in needof furtherwork. It isclear that additionalpsychoacoustical data and suitable technical innovations can alleviate someof the remainingproblems. Others, though, may be better treatedby standardization.
9:15
A3. RODS---An advanced approach to electronic reverberation enhancement. Peter W. Barnett (AMS Acoustics,Ltd., 52 ChaseSide Southgate, London N14 SPA, United Kingdom) This paperwill brieflyreview the recenthistory of reverberationenhancement and then examinethe processesinvolved in thedevelopment of theRODS ( Reverberation-on-DemandSystem) concept. The theory and implementationof the RODS conceptwill be explained,followed by examplesand resultsof recent installations.
S2 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S2 9:45
A4. •plali•tion usingvoice and music •s the source.John Meyer (Meyer SoundLaboratories, 2832 San PabloAvenue. Berkeley, CA 94702) A source-independenttechnique to measureaccurately the amplitude and phase response of soundsystems in concerthalls is discussed. Measurements may be made during live performances or events,using music or voiceas the test signals. Correlation is shownbetween the impulse •esponse and the results obtained using musicsignals. An equalizerthat corrects for manyroom resonances in both amplitude and phase simultaneous- ly hasbeen developed. The effectof thisequalizer on concertsystems in an existingvenue is shown.
10:15
A5. A spatial soundprocessor for headphoneand Ioudsl•eakerreproduction. William L. Martens, Gary S. Kendall,and Martin D. Wilde (SoundSphere,Inc., 1030Davis Street, Evanston, IL 60201)
A spatialsound processor for stereohcadpbone and loudspeakerreproduction is describedthat can posi- tion soundelements within a threc-dimensionalreverberant space surrounding the listener.Spatial motion of soundsources in three dimensionsis createdby dynamicfiltering based on head-relatedtransfer functions. Additionalfilters and delaylines capture air absorptionand Dopplershifting as the propagationtime is manipulatedfor bothdirect and indirect sound. The spatiotemporaldistribution of earlyreflections is captured for a givensource/listener orientation: The gain,delay, and directionalfiltering of simulatedreflections are responsiveto changesin thespecified position and orientation of thesound source and the listener'shead in the simulatedenvironment. The spatialprocessor can be used for headphonereproduction using a head-tracking device,and can also be usedin moretypical reproduction settings such as living rooms with stereoloudspeak- ers.In the lattercase, additional processing is employedto stabilizethe stereoimage and produce a spatially diffusereverberant surround effect over a wide rangeof listeningpositions.
10:45-11.'00
Break
Contributed Papers
11:00 11:15
A6. Binaural simulationtechnique for scale modeling.N. Xiang and J. Blauert(Lehrstuhl far AllgemeineElektroteehnik und Akustik, Ruhr A7. On the model order for the identification of acoustical systems. Universit•it,D-4630 Bochum 1, FederalRepublic of Germany) Nobuo Koizumi (NTT Human Interface Laboratories, Tokyo 180, Japan}and Richard H. Lyon (Departmentof MechanicalEngineering, Roomsimulation for thepurpose of predictingacoustic behavior and MIT, Cambridge,MA 02139) qualityhas recently become a populartopic in roomacoustics. At the Ruhr University,the experience of thebinaural human listener in a room simulationusing both physical and computer modeling has been authenti- The pole/zeromodel and the finite impulse response (FIR) modelare callyrecreated. This paper reports on thelatest stage of thepresent work usedas systemmodels for the identificationof unknownacoustical sys- in binauralroom simulation using a scaleddown physical model, where a tems.The size of the model is particularlyimportant in discrete-time sensitiveminiature dummy head with accuratclyscaled pinnae (scale implementationas it determinesthe convergencerate of adaptationand factor1:10) serves as a receiverto pickup themodel sound field. A versa- capacityof real-time processing. The order of the pole/zero model is relat- tile PC-basedsystem measures the binauralimpulse responses in model ed to modal distributionof systemswhile the order of the FIR model spaceaccording to the m-sequencetransform and alsoearfica out the dependson its dampingfactor. Effective orders of bothmodels are esti- convolutionof theroom-impulse responses with ancchoicspeech or music matedfrom the statisticalproperties of acousticalsystems. In a three- signals.An especiallywide broadbandultrasonic transmitter system is dimensionalenclosure, its volume and reverberationtime are used for necessaryto providethe desiredcomplex stimulation of the soundfield. estimation.It isshown that, whenmodal density of the systemis low, such After the necessarysignal processing, the resultingsignals can bc listened asin a smallenclosure, and frequencyrange is narrow,pole/zero model- to binaurally via headphones.[ Work supportedin part by Deutsche ing can greatlyreduce the modelorder. [Work supportedby NTT Hu- Forsehungsgemeinschaft.] man InterfaceLaboratories, Tokyo 180,Japan.]
11:30-12:00
Panel Discussion
S3 J. Acoust.Soc. Am. Suppl.1, VoL86, Fall 1989 118th Meeting:Acoustical Society of America S3 TUESDAY MORNING, 28 NOVEMBER 1989 ST. LOUIS BALLROOM G, 8:00 A.M. TO 12:00 NOON
SessionB. PhysicalAcoustics I: Scattering,Propagation, Diffraction, and Reflection
JulianD. Maynard, Chairman DepartmentofPhysics, Pennsylvania StateUniversity, University Park, Pennsylvan ia16802
ContributedPapers
8:00 Kristenssonand S. Strrm, J. Acoust.Soc. Am. 64, 917-936 ( 1978); G. S. Sammelmannand R. H. Hackman, J. Acoust. $oc. Am. 82, 324-336 B1. Consequencesof the acoustic bipolar couplings to largeaspect ratio ( 1987) ]. Thesearticles consider scatterers in nonattenuating,liquid me- scatterers.Roger H. Hackman(Physical Acoustics Branch, Naval dia.When objects buried in sedimentare considered, however, sediment CoastalSystems Center, Panama City, FL 32407) attenuationintroduced fundamentally new features into the description It hasbeen suggested that a planewave axially incident on a large of thescattering process. The interactionof an incidentplane wave with aspectratio scatterer couples to boththe insonified end of thescatterer the sediment-seawaterinterface distorts the phase-amplituderelation of andto the end in thegeometric shadow, at lowkD/2 (Dis eitherthe minor the wave incidenton a scattererin the sediment.The descriptionof the axisof a spheroidor the diameterof a cylinder)[Williams et al., J. scatteringof theseinhomogeneous waves requires the introductionof Acoust.Soc. Am. 85, 2372-2377 (1989) ]. This effectis concretelyestab- new, "suitablymodified" basis states for the solutionof the scattering lished,and its experimental consequences arediscussed in some detail. A problem.The modification of the T-matrix and waveguide theories devel- novelclosed-form expression is derivedfor t•e axisymmetric,elastic re- opedat theNaval Coastal Systems Center is discussedand a numerical sponseof a largeaspect ratio target that directly incorporates the bipolar studyof thescattering by a thinspherical shell buried in an attenuating couplingof theacoustic and elastic fields. The form of thefinal expression liquidsediment is presented. is,in somerespects, similar to thatobtained from the generalized geomet- ric theoryof diffractionin thehigh-frequency limit for elasticspheres and cylinders,although the physical assumptions are quite different. Simple, approximateestimates of theparameters involved in thisexpression are 8:45 obtainedand comparedwith a T-matrixcalculation. B4. High-frequencyacoustic scattering from a doublyperiodic ellipsoidal surface: Neumann boundary conditions. Garner C. Bishop and Judy Smith(Naval UnderwaterSystems Center, Newport, RI 02840) 8:15 An exactsolution is obtainedfor the scatterof an acousticplane wave B2. The acousticscattering by a submerged,elastic spherical shell: The from an infinitesurface constructed from a doublyperiodic array of infi- transitionfrom thin to thick shells.Gary S. Sammelmannand Roger niteand parallel elliptical semicylinders. Neumann boundary conditions H. Hackman(Physical Acoustics Branch, Naval CoastalSystems areimposed and the Helmholtz-Kirchhoff integral used to calculatethe Center, PanamaCity, FL 32407) scatteredpressure field. Fredholm integral equations of thefirst and sec- ond kind are used to calculate the surface field. Numerical calculations A fundamentallyoriented analysis of the pole structure of the acoustic are performedto determinethe effectsof geometricparameters on the scatteringmatrix in the low- to high-frequency region (0 S4 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S4 productexpansion. The comparison also shows that omission of theterm sentationfor source-excited-time-harmonic pressurefields in the presence f(•.,i canintroduce substantial errors for x betweenX,•and X,, 2. [Work of a thin elasticspherical shell immersed in differentinterior and exterior supportedby ONR. ] fluids.The newformulation identifies directly the traveling waves with theirmultiple encirclements of the shell by extending the 0 domainfrom itsconventional 0<0•<•r range with periodicity constraints into the un- bounded(multisheeted) domain without theseconstraints. The new for- mulationalso systematizes thetreatment of spherical caps and other trun- catedsections. Periodicity for theclosed shell is recovered by summing 9:15 contributionsfrom an infinite array of imagesources located in the"non- B6. Lips causticsin light backscatteredfrom an acousticallylevitated physical"portion 0 < 0 and0 > rrof theangular space. The rigorous solu- sl•heroidalwater drop. Harry J. Simpsonand Philip L. Marston tion is obtainedby synthesisover a complexspectral continuum, and (Departmentof Physics,Washington State University, Pullman, WA variousalternative representations are derived from it. Specialattention is 99164-2814) givento rigoroushigh-frequency asymptotic forms that describe the wave phenomenain terms of incident and geometrically reflected ray fields, and Cuspdiffraction catastrophes that open up roughly transverse to the alsoin termsof surfaceguided ray fields. The latter are excited by phase propagationdirection are known to existfor bothlight and sound [P. L. matchingof theincident ray field to thetraveling wave modes in theshell, Marston,J. Acoust.Soc. Am. 81,226-232 (1987)]. They may be pro- andthey reach the observer by phase matched detachment [ see, also, P. L. ducedby reflectinghigh-frequency sound from curvedsurfaces. In the Marston,J. Acoust.Soc. Am. 83, 25-37 (1988)]. The phasematching presentresearch, a closely related caustic in whichthe cusp curves join to appliesto directlyexcited leaky waves as well as to wavesthat decay forma pair of lips is studied. This caustic was predicted toexist [J. F. Nye, initiallyinto the fluid; the latter are excited from an exteriorsource by Nature312, 531-532 (1984)] in lightbackscattered from horizontally evanescenttunneling. [Work supportedby the O•ce of Naval Research illuminatedoblate drops of water,provided the axis ratio q = D/H was and David Taylor ResearchCenter. ] withincertain ranges. The associated rays are diffracted at thedrop's surfaceand have only one internal reflection. As q is increasedabove the criticalvalue q4 • 1.31associated with a hyperbolicumbilic focal section, thecusp points at theends of thelips caustic were predicted to mergein the backwarddirection at a "lips event"when q = qL•1.42. For qL• < q < qc2 • 1.58,no causticsfor thisclass of ray areexpected while a 10:f10 secondlips event occurs at qL2- Observationsof farfield scattering from levitateddrops support Nye's analysisand illustratea mechanismfor B9. The opticalillusion of sound.Bradford L. Swartz (Communication producinglips caustics.The observedbackscattering is weak for q Disorders,Central Michigan University, Mt. Pleasant,MI 48859) betweenqL • andq•2. [Work supportedby ONR. ] The graphicdepiction of soundis nota straightforwardmatter. Sta- tionaryrepresentations using lines, waves, or circlesleave the notionof movement(either particle or energy)and pressure fluctuations within the waveto thereader's imagination. The commonsinuous wave shape evi- dentin mostall sourcesof studyoriginates from Lissajous,and resulted 9:30 fromhis transfer of tuningfork motions into a pointof reflectedlight that traverseda screen.Lissajous performed his experimentswith finelyfo- B7. Quasiperiodof variationsin the backscatteringand total cross cusedelectric lamp light, tuning forks, and mirrors. Similar experiments sectionsof sphericalshells. Steven G. Kargl and PhilipL. Marston canbe replicated today using laser light and audio speakers. Thus one can (Departmentof Physics,Washington State University, Pullman, WA morereadily gain an appreciation for thefamiliar wave pattern represent- 99164-2814) ing sound,and seeby the simpleinstruments used to createit, how this waveformis not a truedepiction of theshape of soundwaves in a medium. Quantitativeray methods for leakysurface elastic waves are applied It isin realitya displacementversus time graph. By using multiple forks or [P. L. Marston,J. Acoust.Soc. Am. 83, 25-37 (1988)] to approximate speakersin opposingplanes and tuned to variousfrequencies, the creation thequasiperiod of structures in backscatteringand total scattering cross of the single-and multiple-ellipse Lissajous patterns that are oftenpro- sections.The relevantamplitudes in thismodel contain a Fabry-Perot ducedon an oscilloscopescreen can alsobe accomplished. resonancedenominator, which for spheresis [1 +exp( 2tr/3• + t2rac/c•) ] for the lth classof SEW;x = ka = •oa/cand c• (x) and /3,(x) arethe SEW phase velocity and radiation damping coe•cient. The incrementin x betweenresonances fora givenl followsfrom the condition A(xc/c•)= 1. From the Taylor expansion[c,(x+ Ax)]-• •c•(x) • + Axd(c••)/dx, the approximateresonance spacing Ax•c•,/c forweakly dispersive SEW having a group velocity c•(x) is 10:15 obtained.(This approximation has been noted by others.) The quasiper- iodof a broad structure inthe total scattering cross section isalso approxi- B10. Acoustic surface shape resonances. Alexei A. Maradudin (Departmentof Physics,University of California,Irvine, CA 92717) matedas Ax• • 2•r[ 2 cos0, - (c/c,)( •r - 20•) ] •, sin0, = c/c,,when dispersionisneglected. This follows from a conditionon the phase incre- Acousticsurface shape resonances are vibrational excitations that are mentof the forwardamplitude. Both typesof variationsdue to Lamb localizedin thevicinity of an isolatedprotuberance or indentationon the wavecontributions can be seen in exactand ray-synthesis calculations for otherwiseplanar, stress-free surface of a semi-infiniteelastic medium. The a sphericalshell. [Work supportedby ONR. ] protuberancemay be fabricatedfrom the samematerial as the substrate, or froma differentmaterial. In general,there is an infinitenumber of such resonancesassociated with a givensurface perturbation. Their frequen- ciesare discrete because of theloss of translationalsymmetry caused by thesurface perturbation; they are complex because they overlap the range of frequenciesallowed the vibrations of thesubstrate, into which they can 9:45 decay;and they depend on theshape of the protuberanceor indentation, B8. New rigorous and ray-acoustictraveling wave formulations for andon the relationof the materialproperties of the protuberance(mass source-excited thin elastic spherical shells immersed in fluids. L. density,elastic moduli) to thoseof the substrate.Methods for calculating B. Felsenand J. M. Ho (Departmentof ElectricalEngineering and the frequenciesof the acousticsurface shape resonances associated with ComputerScience, Polytechnic University, Farmingdale, NY 11735) protuberancesor indentationsof simpleforms are described.Results are presentedthat suggest that acoustic surface shape resonances can be stud- A continuousLegendr6 transform spanning the polarangle (0) do- iedexperimentally by thescattering of acousticwaves from thestructure main - o• < •9< o• is employedto derivea rigorousnew integral repre- supportingthem. S5 J.Acoust. Soc. Am. SoppI. 1, Vol. @6, Fall 1909 118thMeeting: Acoustical Society of America $5 10:30 11:15 Bll. Diffractionofa pointsource about a disk.J.J. Dlubac,C.W. Jiang, B14. Predictionof bendingresonances on elasticspheroids due to oblique and B. Yim (David Taylor ResearchCenter, Bethesda,MD 20084- incidentplane waves from exact beam theory and comparison to T-matrix 5000) calculations.M. F. Werby (NORDA, NumericalModeling, Stennis SpaceCenter, MS 39529)and G. C. Gaunaurd(Naval Surface Warfare Computationsof diffractionof an acousticpoint source about a hard Center,White Oak Laboratory,Silver Spring,MD 20903) andsoft disk are presented. The solution, given as a seriesof oblatespher- oidalwavefunctions, isbased on the classical analysis [ J. J. Bowman,T. B. Exactbeam theory predicts the presence of "bending"or flexuralre- A. Senior,and P. L. E. Uslenghi,Electromagnetic and,qcoustic Scattering sonancesfrom oblique incident plane waves. In an earlierpaper, Timo- by$imple Shapes (North-Holland, Amsterdam, The Netherlands,1969), shenkotheory, which assumes a constantmoment of inertiaalong the Chap. 14]. In the limit as the sourcegoes to infinity,the plane-wave beam,was employed to derivean approximatetheory to predictflexural solutionis recovered. Computations with thefield point on thedisk verify resonances.Good agreement was found (for thelower modes) with the the boundaryconditions. Parameter studies are presented,and conver- exact T-matrix calculationsbased on exact3-D elastodynamics.It is pos- gencecriteria are discussed. sibleto solve the exact beam equation using phase matching techniques to predict"bending" resonance locations. Such calculations are performed for spheroidsfor aspectratios from 1.5to 10,with extremely good agree- mentbetween beam theory and the exact T-matrixcalculations for all modes. 10:45 B12. Reflectivity of monolayersof bubblelikescatterers at water/solid 11:30 interfaces.Ivan Tolstoy (Knockvennic,Castle Douglas DG7 3PA, SW Scotland) B15. Soundscattering from a submergedfinite cylinderwith ribs. Philip J. Moser,Jian-Ren Yuan, and Herbert•berall (Departmentof By redistributingthe energyflux of an incidentplane wave, a toorio- Physics,Catholic University of America,Washington, DC 20064) layerof identicalbubblelike scatterers at an interfacemay, at frequencies closeto the monopolcresonance tot,, drastically alter the reflectivityof the The effectsof an almost-periodicallyspaced set of ribson the scatter- surface.Reflectivity calculations arc given for a waveincident normally ing of soundfrom a submergedcylinder of finitelength are investigated. uponinfinite square lattices (of basisl) for threemodels: ( 1) air bubbles The cylinderis eitherassumed soft (corresponding to a verythin air-filled in a liquidfullspace, (2)' airbubbles at a water/hardsurface or water/ shell)or rigid (correspondingto a heavythick shell), and its scattering elasticplate interface, and (3) air-filledcavities in a rubberlayer at such amplitudeis obtained using the geometrical theory of diffraction.The rib an interface.These illustrate the essentialrole of multiple scatterand, scatteringamplitude is found from curvature corrected results of thetheo- mostinterestingly, the possibility of reflectivitynulls for incidentfrequen- ry for a flat platecarrying a rib. Numericalresults are shown for thecase ciesto-•too- For case(3), useof rubberconstants given by earlierwriters of softcylinders; they exhibit the onsetand emergenceof rib effectsover [Gaunaurdetal., J. Acoust.Soc. Am. 65, 573-594 (1979) ] showsthat it the backgroundof scatteringfrom the bodyof the cylinder. is theoreticallypossible to selectvalues of kl suchthat a hardwall (or an elasticplate) becomesfully anechoicfor given to•-too.For the rubber typesconsidered, the effective bandwidth of thiseffect (99% anechoicity) ,xco/coo varies between 0.01 and0.05. [Work supportedby ONR.] 11:45 B16. Diffraction tomographieinterferometry, a new imaging technique. Woon S. Gan (Acoustical ServicesPte Ltd., 29 Telok Ayer Street, Singapore0104, Republicof Singapore) A new imagingtechnique, diffraction tomographic interferometry 11:00 (DTI), is proposed.Double-exposure DTI and time-averagedDTI are considered.The Bornapproximation is usedto obtainthe scatteredwave B13. Acoustical imagesof a submergedaluminum sphere. Char]cs F. Gaumond,Brian H. Houston(Codes 5132 and 5136,Naval Research and the filtered-backpropagationalgorithm in the reconstruction.For double-exposureDTI, summationof the two tomograms gives the diffrac- Laboratory,Washington, DC 20375-5000),and SheilaWoo (Sachs Freeman Associates,Landover, MD 20785) tion tomographicinterferogram. Reconstruction gives a cross-sectional imagecovered with interferencefringes. For time-averagedDTI, the in- Acousticalimages using reflection tomography [P. B. Abrahamand tensityof thereconstructed image is expressed in termsof thecharacteris- C. F. Gaumond, J. Acoust. Soc. Am. 82, 1303-1314 (1987)] wcrc made tic function.This givesa cross-sectionalcontoured image of the object from 3-D acousticalscattering data. The data were taken from a 7.6-cm modulatedby a systemof interferencefringes. For the proposedexperi- (3.0-in.) solidaluminum sphere using a semicircularhoop array with mentalsetup for thedouble-exposure DTI, theinterferogram of theob- radius0.85 m, which rotatedaround the sphere.The clementspacing in ject,a turbineblade in its unstressedstate and stressed state, is recorded the hoopwas 2 ø. The hoopwas rotated in 2ø increments from the back- with the objectilluminated from variousdirections. For the proposed scattereddirection to the forwarddirection. The frequencyrange of the experimentalsetup for thetime-averaged DTI, a diffractiontomogram is data was 2.7-25 kHz. Theseacoustical images were comparedwith ana- exposedfor a periodof timeby rotatingthe illuminatingsource around lytically producedacoustical images. the objectseveral times, during which the objectexecutes a motion. S6 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S6 TUESDAY MORNING, 28 NOVEMBER 1989 ST. LOUIS BALLROOM H, 8:00A.M. TO 12:15P.M. Session C. Underwater Acoustics I: Acoustic Fields John Perkins, Chairman Code5160, Naval Research Laboratory, Washington, DC 20375 Chairman's Introduction--8:00 ContributedPapers 8:05 Can matchedfield processingbe usedto determinethe sound-speed environmentgiven a knownsource? Recent computations are very prom- CI. Three-dimensionalmatched-field processing with an eigenvalue isingin spiteof expecteddifficulties relating to the nonuniquenessof solu- method.Michael D. Collins,W. A. Kuperman,John S. Perkins,Laurie tions.In the range-independentcase, the essenceof the approachis to find E. Tinker (Naval ResearchLaboratory, Washington, DC 20375), and "the" sound-speedprofile that will maximizethe powerseen at the array. John Glattetre (NorwegianDefense Research Establishment, P.O. Box Thishas resulted in a profilethat is "close to" thetrue profile and found by 115, N-3191 Hotten, Norway) meansof an exhaustivesearch of the parameterspace describing the fam- ily of potentialprofiles. The parameterspace consists of thecoefficients of An eigenvaluemethod for three-dimensionalmatched-field process- the major empiricalorthogonal functions (EOFs) computedfrom the ing with a verticalarray has been developed for localizationof multiple covariancematrix of sound-speedprofile samples. Subsurfaces of the pa- sourcesin rangeand depth (bearingcan also be determinedwith the rameter spaceare easily plotted and show well-defined,informative methodif theenvironment isasymmetric). The method involves applying trends.Unfortunately, iterative techniques investigated so far haveshown the Bartlettbeamformer to the eigenvectorsof the cross-spectraldensity convergencedifficulties in that they findlocal maxima which correspond matrix. This approachappears to be more effectivethan the standard to profileswhich are not closeto the true profile.Results will showthe eigenvalueapproach involving beamforming on a linearcombination of proceduresensitivity as a functionof arraylength, source range, depth, severaleigenvectors. Eigenvalue beamforming methods are basedon the andfrequency. The keyto successfor thetechnique is thecomputation of orthogonaltargets approximation [R. F. Oragg,NRL Rep. 9143, Naval "good" EOFs. ResearchLaboratory, Washington, DC (1989) l- The spatialcorrelation over a horizontalarray of two sourcesis small providedthe sourcesare sufficientlyseparated in bearing.On the other hand,the spatialcorrela- tionover a verticalarray of two pointsources may be relativelylarge for somesource locations (i.e., the Bartlett matched-fieldprocessor for a 8:50 verticalarray has relatively large sidelobes). However, simulations sug- gestthat thiseffect does not seriouslydegrade the eigenvalueprocessing C4. Low-frequencyacoustic tomography using matched field processing. methodfor verticalarrays. Source motion will furtherenhance this pro- II. A. Tolstoy (Code 5120, Naval ResearchLaboratory Washington, cessingmethod. DC 20375) Part I discussedthe difficultiesand successesof the MFP techniquein a range-independentenvironment. Part II discussesprogress to data (sim- ulationsonly) concerninga range-dependenteddy environment. First, a 8:20 sound-speedstructure is computedas the sumof two Gaussianeddys. Next samplesound-speed profiles are computedalong a selectedradial C2. A perturbationmethod for normalmode theory with variablevelocity and used to calculate the basic EOFs for the entire environment. The profile. M. F. Werby (Schoolof Physicsand Astronomy,University of EOFs themselvesare fairly insensitiveto how the samplingis doneal- Minnesota,Minneapolis, MN 55455and NORDA, NumericalModeling thoughtheir coefficients can change significantly. Given a setof EOFs, StennisSpace Center, MS 39529},and Paul Ellis (Schoolof Physicsand plotsof theircoefficients as a functionof rangeand cross range show well- Astronomy,University of Minnesota,Minneapolis, MN 55455) defined,easily interpolated behavior. Moreover, only two EOFs are need- Normalmode theory is a verywell-developed technique for predicting ed to describea sound-speedprofile to a highdegree of accuracy.Results the propagationloss in the oceanfor most stratifiedenvironments and will be presentedexamining the sensitivityof MFP to changesin the velocityprofiles, as well as somerange-dependent environments. So why coefficients.Regions of high sensitivityrequire frequent profile estima- another method? The reason is that sometimes it is desirable to retain the tionsand increasedenvironmental sampling. Progress to date suggests simpletrigonometric functions obtained from constantvelocity profiles. that the largefamily of parametervalues (at leasttwo parameters,i.e., This is usuallydone by employingthe approximate scheme of usingmulti- EOF coefficients,per profile) plus iterativedifficulties as seenin Part I ple layersin eachofwhich the velocityprofile is assumedconstant. In this mayrequire the application of computationallyintensive techniques such work, it is shownthat this can be carriedout nicelyby representingthe as simulatedannealing. velocity profile in terms of a mini-max fil usingTschebyscheff polynomi- alsand a first-orderperturbation technique derived from Sturm-Liouville theory.The derivationleads to a compactclosed expression that is easily programmable.The resultsfrom the formulation are compared with other methods. 9:05 C5. Environmental mismatehingeffects on sourcelocalization processing in mode space. E. C. Shang and Y. Y. Wang {CIRES, University of Colorado/NOAA, Wave PropagationLaboratory, Boulder, CO 80303) 8:35 The performancedegradation of the matchedfield processing (MFP) causedby environmentalmismatching has been an intriguingsubject. A C3. Low-frequencyacoustic tomography using matched field processing. greatdeal of studyhas been done by numericalsimulations. In thispaper, I. A. Tolstoy(Code 5120, Naval ResearchLaboratory, Washington, this problemis studiedanalytically. Analytic investigationis important DC 20375) becauseit revealssome of the physicalinsights and providesguidelines for S7 J. Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S7 a "self-coherent"scheme [H. Bucker,J. Acoust.Soc. Am. Suppl.1 84, 9:50 S19 (1988) ]. It hasbeen illustrated that the sourcerange information is C8. Effect of internal soiltonson shallow-watersound propagation. Ji- associatedwith themodal phase difference and the source depth informa- xun Zhou,Xue-zhen Zhang, and Peter Rogers(School of Mechanical tion isassociated with themodal amplitude ratio whenMFP is performed Engineering,Georgia Institute of Technology,Atlanta, GA 30332) in modespace. Therefore, the modalwavenumber plays a key role in sourcelocalization. Mismatching effects can be investigatedanalytically Therehave been many reports on observationsof naturallyoccurring by takingadvantage of calculatingthe modal wavenumber error by modal internalsolitary wavetrains (soiltons) in the coastzone, especially in the perturbationtheory. [Work supportedby NOAA and NOSC.] summer.The mechanismfor the generationof thesenonlinear internal waveshas been investigated in the geophysicsand fluidmechanics com- munity. Little attention,however, has beenfocused on their effecton soundpropagation with the exceptionof the work of Baxter and Orr, which was basedon ray theory [L. Baxter and M. H. Orr, J. Acoust. Soc. Am. 71, 61-66 ( 1982) ]. In thispaper, the parabolic equation (PE) model 9:20 (IFD or PAREQ code)is usedto numericallysimulate the effect of inter- nal solitonson low-frequencysound propagation in the coastzone. The C6. Nonlinear biasof travel time in modalocean acoustic tomography. E. resultsshow that soundtransmission loss is sensitiveto solironparameters C. Shang(CIRES, Universityof Colorado/NOAA, WavePropagation (suchas wavelength, amplitude, etc. ) andwave packet parameters ( posi- Laboratory,Boulder, CO 80303) tion, number,and propagationdirection). The resultscan be usedto ex- In a previouspaper [ E. C. Shang,J. Acoust.Soc. Am. 85, 1531-1537 plain,at leastpartly, an interestingexperiment phenomena: The frequen- (1989) ], oceanacoustic tomography based on theadiabatic normal mode cy responseof shallow-watersound propagation under the conditionof a theoryhas been developed. Mesoscale structure can be inverted by usinga thermoclineduring the summeris often a strongfunction of spaceand linearrelationship connecting the modal travel time perturbation and the time,and soundpropagation over certain frequency ranges has an abnor- sound-speedperturbation. This linearrelationship is only an approxima- mal attenuationthat cannotbe explained by usingconventional models of tion dueto the followingtwo facts:( 1) neglectingthe differencebetween soundpropagation. [Work supportedby ONR. ] perturbedmodal eigenfunction and the background model eigenfunction; (2) takingthe first-order expansion ofthe "potential function" AK: (z,r) to calculatethe modalwavenumber perturbation. Comparing with ray tomography,item( 1 ) is theterm corresponding to theray pathperturba- 10:05 tion. It has beenfound that item (1) is not significantand the problem C9. Measurement of shallow-water sound transmission on the New causedby item (2) can be recoveredby a "post-correction"procedure Jersey continental shelf. Lynne Maiocco, William Carey (Naval proposedin thispaper. [Work supportedby NOAA and ONR.] UnderwaterSystems Center, New London,CT 06320), Edwin Parssinen (Kildare Corporation,95 Trumbull Street,Suite D, New London,CT 06320),and James Doutt (WoodsHoleOceanographicInstitute,Woods Hole, MA 02543) Calibrated acoustic transmission measurements were made on the 9:35 New Jerseycontinental shelf in the vicinityof the AMCOR 6010 bore- C7. The normal mode theory with bottom elasticity and range hole,an areawith knowngeophysical properties. The waterdepth was 70 dependence.Juan I. Arvelo (Code U25, Naval SurfaceWarfare Center, m andthe experimentwas conducted under known oceanographic condi- tionsas shown by measurementsof salinity,temperature, and soundspeed 10901New HampshireAvenue, Silver Spring,MD 20903-5000)and Herbert0befall (Departmentof Physics,Catholic University of versusdepth. The acousticmeasurements were performed with a vertical America, Washington,DC 20064) arrayof 24 equallyspaced hydrophones, one of whichwas on thebottom. A calibratedsource was used to obtaintransmission loss data to rangesof The theoryof normalmodes has been extended to accommodatethe 26 km at discretefrequencies from 50 to 600 Hz. The data from three effectsof shearwaves in the elasticocean floor. The effectof absorption hydrophoneswere processed for a 4-km rangeusing techniques developed has alsobeen included as the imaginarycomponent of the shearand by Frisk and Lynch [J. Acoust.Soc. Am. 76, 205-216 (1984)] to yield compressionalwavenumbers. A semi-infiniteelastic bottom basement wavenumberspectra. The transmissionresults of threehydrophones at 50 layer hasbeen included and it was foundthat the absorptioncauses the Hz werefound to havean interferencepattern clearly resulting from two wavenumberspectrum of the radiatingmodes to be inherentlydiscrete, modes.Low-frequency data (20 to 50 Hz) wereobtained at discreteloca- hence,the numberof radiatingmodes is drasticallyreduced relative to the tionsfrom the buoyto yieldvalues of normalboundary impedance. Trans- onesobtained with the deepfalse boundary. The rangedependence of the missionloss data were obtained at eightfrequencies, 50 to 600 Hz, alonga acousticproperties and of the boundariesof the oceanhas been included constantdepth and gradually changing depth radial. The constantdepth by a modifiedversion of the adiabaticnormal mode theory. Comparisons radial resultswere compared to calculationsperformed with IFD-PE and are made with experimentalmeasurements of the transmissionloss of SAFARI usinga geoacousticmodel based on measuredgeophysical prop- underwaterexplosives and with resultsfrom Collins' HEPE mode. erties.Predicted and measuredlevels generally agreed, however, differ- [Work supportedby the IndependentResearch Board of the Naval Sur- encesin thecomputed and themeasured modal interference patterns were faceWarfare Center and by NORDA. ] observed. 10:20-10:30 Break 10:30 This paperwill describethe resultsof a modelscale experiment on soundpropagation in a penetrablethree-dimensional wedge with a fluid C10. Experimental results of soundpropagation in a penetrable wedge bottom. The wedgemodel is 1.2X2.4 m and has a Mylar-kevlar mem- with a fl.uidbottom. Stewart A. L. Glegg and Ian House (Center for branethat separatestwo fluids (water and ethylenegylcol) with sound Acousticsand Vibration, Department of Ocean Engineering,Florida speedsof 1500m/s alad1700 m/s. The modelis suspendedin a largewater Atlantic University, Boca Raton, FL 33431) tank so that reflectionsfrom the edgesof the model are not significant. S8 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $8 Theexperiment has concentrated on acrossslope propagation at several modespace matched-field techniques more applicable to actualphysical differentdownslope ranges, using a high-frequencypulse. The pulse spec- environments[T. C. Yang,J. Acoust.Soc. Am. 82, 1736-1745( 1987}]. tra havebeen compared with Buckingham'stheoretical model for sound A freshapproach is broughtto thisproblem here using computer simula- propagationin a penetrablewedge. [Work supportedby ONR.] tionsof matched-fieldprocessing in a Pekeriswaveguide to demonstratea new methodfor extractingmodal amplitudes from data.These simula- tionsdemonstrate that thisnew method of modalamplitude extraction allowssuccessful mode space matched-field detection and localization at array aperturessignificantly smaller than thoseallowed by previously 10:45 usedalgorithms. Finally, the algorithmwill be discussedin the general C11. The resolutionof modal Doppler shifts in an acousticwaveguide. contextof dimensionalreduction techniques. Hee Chun Song and Arthur B. Baggeroer(Department of Ocean Engineering,Massachusetts Institute of Technology,Cambridge, MA 02139) A methodfor estimatingthe sourc•velocity from the acousticsignal 11:30 fieldpropagating in a dispersiveoceanic waveguide is presented.The sig- nal is observedat an omnidirectionalreceiver in the presenceof additive C14. Combinedoceanographic and acousticmodeling. Michael Porter, whitenoise. The sourcetransmits a continuous-wave(cw) signal,and the Steve Piacsek, Laurel Henderson, and Finn Jensen (SACLANT source motion is assumedto be uniform (unaccelerated). The ocean is Undersea,Research Centre, I-19026 La Spezia,Italy) modeledas a waveguidethat is horizontallystratified with an arbitrary sound-speedprofile in thevertical. The acousticfield generated by a mov- An importantclass of oceanographicmodels is themixed layer models ing pointsource in termsof normalmodes predicts that eachmode has a that, simplystated, provided predictions of mixedlayer depthand tem- differentDoppler shift, whichcontrasts with a singleDoppler shift in a perature.These predictions take into accountthe surface winds that drive homogeneousunbounded medium. The modesare consideredto remain the mixingand the solarradiation that heatsthe mixedlayer. As interest phaselocked with eachother, i.e., coherent.The sourcevelocity, includ- has increasedin usingsuch models as a stepin the processfor making ing the numberof modesinvolved, can be estimatedthrough resolving acousticforecasts or nowcasts,a recurringquestion has been that of what informationis neededfor accurateacoustic predictions. For instance,how Dopplershifts. The methodis a time-domaininterpretation of a recently preciselydo mixed layer depthsneed to be calculated?!s mixed layer developedeigcnstructure technique for multitargetdirection finding with passiveantenna arrays, in conjunctionwith smoothingpreproecssing temperatureimportant? These and other questions are addressed through schemeto deal with coherentmodes [Shan et aL, IEEE Trans. Acoust. a parameterstudy using synthetic sound-speed profiles as input to an SpeechSignal Process. ASSP-33 (4) (1985)]. Simulationresults that acousticmodel. The implicationsare demonstrated in a somewhatmore illustratethe performanceof the methodare presented. concreteform in a two-stepprocess: The SSPis predicted using particular mixedlayer models (Niiler, Mellor-Yamada,Garwood, etc. ) andpassed toan acoustic model to providetransmission loss calculations. The results are alsocompared to predictionsbased on both measuredand historical SSP measurements. 11:00 c12. Impulse responsefor propagation in the ocean sound channel. StanleyM. Flattr, John Colpsi,Timothy F. Duda, Galina Rovner,and Jan Martin (PhysicsDepartment, University of California,Santa Cruz, CA 95064) 11:45 Usingthe SDSC/CRAY and the parabolicequation, pulse propaga- C15. The transportcoherence of acousticfields as a criterion for range tion wassimulated in the oceansound channel. Pulses with 200 frequen- dependencein an ocean channel. B. G. Hurdle (Naval Research ciesbetween 10 and 50 Hz weresynthesized, giving a pulsewidth of ap- Laboratory,Code 5103, Washington,DC 20375) proximately25 ms, and a pulserepetition time of 5 s. The wave fronts createdby thistechnique have a simpleand intuitive character, unlike the A paperwas presentedlB. Hurdle, J. Acoust.Soc. Am. 55, 452 fieldsfor singlefrequencies. Pictures of wave frontsout to 1000km, and (1974) ] on the natureof acousticinterference fields in a geometricdisper- possiblya movie,will beshown. Wave front reconstruction by meansof a siveocean channel. The structureand multiplecomponents of the fieldsas muchfaster calculation in termsof ray theorywill be demonstrated.The a functionof the velocityof the source,the geometryof the boundaries, effects of internal waves on the wave fronts will be described in terms of andthe sound-speedstructure of thechannel were discussed. This paper wavefront corrugations, and the useof impulsemeasurements over verti- reviews the nature of the fields and, based on the field characteristics, cal arrays to measurethe propertiesof the internal-wavefield will be providesa criterionfor determiningthe degreeof rangedependence of an discussed.[Work supportedby ONR, Code 1125OA.] oceanchannel. This criterionis basedon the transportcoherence of the field and is measured in terms of the correlation coefficient of the trans- portedfield as a functionof theseparation distance between two receivers asthe fieldis transportedthe distance between them. The differencein the transportcoherence and the coherence in thefield for beamforming will 11:15 be discussed.Some other applicationsof this conceptwill also be dis- cussed. C13. Modal matched-fieldprocessing with smalla•rture verticalarrays. G. B. Smith, G. M. Frichter, IV (Naval Ocean Research and DevelopmentActivily, Code 244, StennisSpace Center, MS 3952% 5004), C. Feuillade(Syntek Engineering and ComputerSystems, Inc., 2101 East JeffersonStreet, Rockville, MD 20852), and C. L. Byrne (Departmentof Mathematies,University of Lowell,Lowell, MA 01854) 12:00 In the past years, much attention has been given to mode space C16. The use of neural networks in acoustic coherence measurements. matched-fieldprocessing in shallow-water,low-frequency environments. Mark J. Beran (Department of Electrical Engineering, Catholic In thistype of situation,mode space matched-field processing has several Universityof America,Washington, DC 20064) advantagesover conventionalmatched-field processing such as: lower procc•ing dimension, better sidelobe rejection, and the ability to dis- The tntensity dlstnhuuon of an incoherent target may he determined criminatesignal against modal noise. However, extraction of modalam- from coherence measurements. It is shown here how a neural network plitudesfrom databy the methodpioneered by Shang[J. Acoust.Soc. may be developedto solvethis inverseproblem. The effectof random Am. 77, 1413-I 418 (1985) ] is not veryreliable for smallaperture arrays. phasevariations resulting from ocean temperature fluctuations is consid- Attemptshave been made to overcomethis problemin orderto make ered in the formulation. S9 d. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S9 TUESDAY MORNING, 28 NOVEMBER 1989 ST. LOUIS BALLROOM B, 8:30 A.M. TO 12:00 NOON SessionD. PsychologicalAcoustics I: Temporal Effects and Binaural Phenomena D. Wesley Grantham, Chairman Bill WilkersonHearing and SpeechCenter, 1114 19thAvenue, South, Nashville, Tennessee 37212 ContributedPapers 8:30 9:00 D1. Comodulationmasking release (CMR) using coherent complex D3. A comparisonof comodulationmasking release and profile analysis. envelopes.Mark P. Haggard and Deborah A. Fantini (MRC Institute Deborah A. Fantini (MRC Institute of Hearing Research,University of HearingResearch, University Park, Nottingham NG7 2RD, England) Park,Nottingham NG7 2RD, England),Gregory P. Schooneveldt,and BrianC.J. Moore(Department ofExperimentalPsychology, University In comodulationmasking release (CMR), thresholdsfor a signal of Cambridge,Cambridge, England) maskedby a narrowband of noiseare improvedwhen a second,flanking, noiseband is added which has a similaramplitude envelope to thatof the Large reductionsin signalthresholds may be observedunder stimulus maskingband. The parametersof "similarity" are many; this study ad- conditionsin whichthere is commoninformation in a numberof auditory dressessharing of frequencycomponents in complexmodulations. The channelsor critical bands.In comodulationmasking release (CMR) followingwas determined: (a) whetherthe auditorysystem may extract tasks,a reductionin the thresholdof a signalin a narrow-bandnoise componentsof complexmodulation in order to achievea CMR; (b) occurswhen a flankingor cueband of noiseis also present, provided that whethera "missingfundamental" of a complexmodulation matches up the flankingband has an amplitudeenvelope which is correlated with that with a present modulation componentto provide a CMR; and (c) of the masker band. This reduction in threshold occurs in dichotic condi- whetherthe differenceor beat frequencyof a complexmodulation may tions,where the flankingband is presentedin the oppositecar to the providea CMR, for bothharmonic and inharmoniccomplexes. The re- signal-plus-masker.In profileanalysis (PA), the overalllevel is rando- sultsindicate a smallbut significantCMR with complexmodulation; the mizedfrom one stimulus to thenext. The thresholdfor detecting an incre- reductionin CMR by comparisonwith sinusoidalmodulation may reflect mentin levelof a singlecomponent is reduced when flanking components modulation-modulationmasking. Insignificant CMRs were obtained equalin amplitudeto thepedestal are added. In general,this effect has not wherecomponents would need to be separatedto registerenvelope simi- beenfound when the flanking components are presented dichotically. The larity betweendiffering complex modulations. presentwork examinesthe extentof the similarity,and divergence,be- tweenCMR and PA. For the PA conditions,a five-componentsinusoidal complexwas used; its levelwas varied randomly over a 40-dBrange. For the CMR conditions,each of the fivecomponents was 100% sinusoidally amplitudemodulated at a 10-Hz rate.Two intermediateconditions of PA without the level variation, and CMR with the level variation, were also run. All four conditionswere run monaurallyand dichotically.A refer- 8:45 encecondition in which only one maskingcomponent was presentwas also used. For PA conditionswhen the stimulus level was randomized, D2. The role of envelopeinformatlon in CMR tasks.Deborah A. Fantini thresholds in the monaural and dichotic conditions were 16 and 11 dB (MRC Institute of Hearing Research,University Park, Nottingham lowerthan in the referencecondition, respectively. By contrast,for the NG7 2RD, England) CMR conditions, whether the stimulus level was randomized or not, In comodulationmasking release (CMR), thresholdsfor a signal thresholdsin the reference,monaural and dichotic conditions differed by masked by a narrow-band noise are reduced when additional noise is little morethan 3 dB;i.e., no CMR wasobserved. Further experiments are now in progressto determinewhy a CMR was not observedin these present,provided that the additionalnoise has the sameamplitude enve- conditions. lope fluctuationsover time as the originalmasking band. Recently,some speculationhas appeared in the literatureas to the specificsource of infor- marionwhich providesthe reductionin maskedthreshold. In the present work, the importanceof changesin the envelopeof the maskerband, due to the additionof the signalat threshold,in providingthe CMR, hasbeen investigated.First, CMR thresholdswere collected in sinusoidallyampli- tude-modulatednoises which varied either in the phaserelationship of their envelopesor in their modulationdepth. Then, envelopediscrimina- 9:15 tion (ED) thresholds were collected for changesin envelope phase dispar- D4. Spectral fusion based on coherenceof amplitude modulation. ity and for changesin modulationdepth per se.The data indicatethat the Stanley Sheftand William A. Yost (Parmly Hearing Institute,Loyola patternsof CMR thresholdswith envelopephase disparity and with mod- University,6525 North SheridanRoad, Chicago,IL 60626) ulation depth vary differentlyfrom that of the ED thresholds.A simula- tion wasthen conducted in which the stimuluswaveforms were processed The abilityof listenersto attendto a subsetof components( the target) througha modelthat includesan auditory filter [equivalentrectangular of a tonalcomplex was investigated using a forced-choiceprocedure. The bandwidth (ERB) filtering], a square-lawnontinearity, and a sliding stimuluswas an eight-componenttonal complex.The componentsof the temporal window [an equivalent rectangular duration (ERD) filter]. target subset (n = 1, 2, or 3 components) were distinguishedfrom the The envelopeswere then extractedfrom the processedwaveforms to de- complexby coherentamplitude modulation (AM). Each trial of the cued terminewhat actualchanges in the envelopeswere occurring in the signal 2IFC task was precededby a presentationof the coherentlymodulated intervalsin both CMR and ED. The resultsof this waveform analysis target componentsin isolation.For each interval of a trial, either the indicatethat discriminabilityof envelopechanges due to the additionof targetor an equalnumber of nontargetcomponents shared the coherent the signalat thresholdin the CMR taskare insufficientto explainCMR. AM. Subjectswere required to detect the interval in which the target S10 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S10 componentswere coherently modulated. The remaining components of Maskingof onemodulation frequency by another was measured in a thecomplex were all either modulated at differentrates, modulated with a 3IFCtask. The carrier was broadband noise at a spectrumlevel of 30dB randomshift of the target-modulator phase angle, or not modulated. Per- SPL.This was amplitude modulated by a sinusoidalmasker or a sinusoi- formantewas measured asa functionofthe number of target components, dalsignal plus a sinusoidalmasker. Masker modulation depth was fixed at theharmonic relationship among the target components, and the har- 25%or 50%,and masker frequencies ranged from 2-800 Hz. Threshold monicrelationship between the targetand nontargetspectral groups. signalmodulation depth was measured for signals of 4, 16,100, and 400 Withcoherent AM of harmonictarget components, the increment in d' Hz.The masking functions showed broad tuning for modulation frequen- with increasingnexceeded predictions based on thecombination of inde- cyconsistent with that seen by Bacon and Grantham [J. Acoust. See. Am. pendentsources of information,suggesting that these stimuli may be pro- 85,2575-2580 (1989) ]. Forexample, for a signalmodulation frequency cessedas an entity by the auditorysystem. [Work supportedby of 100Hz, the amountof maskingonly varied over 5 dB for masker NINCDS. ] modulationfrequencies from 40 to 140 Hz. Thresholds were phase depen- dentfor masker frequencies ofhalf, equal to, or twice the signal frequency. For equalsignal and masker frequency, thresholds were consistent with detectionofdifferences inenvelope power, independent ofsignal/masker 9:30 phase.Results are discussedin termsof modelsof modulationdetection. [Work supportedby grantsfrom NINCDS (NS12125)and NICHHD DS. Sensitivityto envelopecoherence. Virginia M. Richards,David (HD-07151).] M. Green, and Zekiye A. Ohsart (PsycheacousticsLaboratory, Departmentof Psychology,University of Florida,Gainesville, FL 32611) In a 2IFC paradigm,listeners indicated whether two, simultaneously presented,sinusoidally amplitude-modulated tones had envelopesthat werein-phase rather than out-of-phase. The depthof modulation,m, nec- essaryfor 70% correctdecisions was determined.Thresholds were ob- 10:00 tainedfor carrierfrequency separations of 2/3 and4/3 of an octave,for geometricmeans of thetwo carriers ranging from 500-8000 Hz, andfor D7. Transition from simultaneousto forward maskingIII: Phaseeffects. modulationrates varying from 4 to 320 Hz. Increasingthe frequency DanielL. Weber(Department of Psychology,Wright StateUniversity, separationof the two carriersfrom 2/3 to 4/3 of an octavehad little effect, Dayton, OH 45435) reducingsensitivity by at most5 dB. Similarly,thresholds were nearly independentof thegeometric mean of the tonalcarriers. Plotting detec- The relativephase of a sinusoidalmasker and signal necessarily has a tion thresholdas a functionof modulationrate yieldedcurves that were largeeffect upon the signalpower required for a simultaneous-masked threshold but has no effect on the forward-masked threshold. Threshold roughly low passin character:For modulationrates lessthan 100 Hz, performancelevels were approximatelyconstant, and, for modulation for a 20-ms(20-ms raised cosine ramps with no steadystate produced a ratesless than 100 Hz, sensitivityto envelopephase deteriorated with signal20 ms long measured at thehalf-voltage points on the envelope), l- increasedmodulation rate. [Work supportedby the Air ForceOffice of kHz sinusoid,masked by a 300-ms(no ramps),1 -kHz sinusoid,was mea- ScientificResearch and the NationalInstitutes of Health.] suredas a functionof the signal'stemporal position to examinehow the phaseeffect disappeared in the transition from simultaneous to forward- maskingconditions. Although listeners differ, the phase effect remains 9:45 unchangedat leastuntil 10% of thesignal energy is subjectto forward maskingand is substantially reduced, if noteliminated, when 50% of the D6. Phaseeffects in maskingof one modulationfrequency by another. signalis subject to forwardmasking. These results are interpreted to indi- Elizabeth A. Strickland and Neal F. Viemeister (Department of catethat forward masking tends to bethe more sensitive detection process Psychology,University of Minnesota,Minneapolis, MN 55455) whena signalis subjectto bothsimultaneous and forward masking. 10:15-10:30 Break 10:30 showedno effectfor Af= 100 Hz, with thresholdsincreasing with in- creasingonset asynchrony when Afwas 10 or 50 Hz. Theseresults show DS. The effect of onsetasynchronies between components on the ability that onsetasynchrony between components does not diminishthe effects to detect a delayed componentembedded in a dietie complex. M. of spectrallysynthetic binaural processing. [Work supportedby NIH. ] A. Stellmack and R. H. Dye (Parrely Hearing Institute, Loyola University,6525 North SheridanRoad, Chicago,IL 60626) The purposeof thisstudy was to extendthe author's previous research 10:45 on the processesthat produceinterference when observersattempt to lateralizea targetcomponent in the presenceof other components.The D9. Onset coding of spatial information in trains of identical transients. presentstudy examinesthe effectof an onsetasynchrony between an in- Kourosh Saberi and David R. Perrott (PsychoacousticsLaboratory, teraurallydelayed target (IDT) and dietic "distractor"components. CaliforniaState University, Los Angeles,CA 90032) ThresholdAIDTs weremeasured for a 753-Hz tonepresented against a In a studyof the effectsof onsetsin lateralizingbursts of noise,Tobias dioticbackground of six additionalcomponents with Af's of 10, 50, or andSchubert [J. Acoust.Soc. Am. 31, 1595-1605(1959) ] demonstrated 100Hz. Onsetasynchronies between the distractors and target were 0, 50, that, dependingon the durationof the ongoingnoise burst, it takesany- 100,or 200 ms,with the distractorsalways gered on first.The durationof wherebetween 4 to 35 timesas muchonset interaural time disparityto the target was always200 ms. The duration of the distractorswas such offsetongoing disparities. They, therefore,demonstrated that onsetsdo that they remainedon for the duration of the target (all componentswere not necessarilyprovide the major cue to auditory localization.In the pres- geredoff simultaneously).The rise/decaytime of all componentswas 10 ent study,onset and ongoinginformation in a lateralizationparadigm is msand each component was presented at 55 dB SPL.The largestthresh- set in spatialconflict to determinethe amount of onsetinteraural differ- oldswere obtainedwhen /Xfwas 50 Hz at eachonset asynchrony. One enceof time (IDT) necessaryto offsetongoing IDTs of transienttrains of subjectshowed no effect of onset asynchronyand the secondsubject variousdurations. The majordifference between this study and the Tobias Sll J. Acoust.Soc. Am. Suppl. 1, Vol.86,.Fall 1989 118thMeeting: Acoustical Society of America Sll and Schubertstudy was that this studyused identical transients [e.g., surroundingthe probeshowed that, as the noisewas displaced from the interclickinterval (ICI) = 1.5to 2 mscorresponding to a 500-to 750-Hz 45ø left to the 30ø left position,the decrease of thenoise level in theleft ear tone] and not noisebursts. Unlike the Tobiasand Schubertstudy, it was wasless important than the increase in theright ear. Addingthis variable- found in this studythat a singledichotic click could easilylateralize a level noiseto the fixed signalshould thus result in monauralmasked contralateraltransient train of upto 200clicks; producing the sensation of thresholdsthat are not appreciably different in the left (i.e., thenear) ear a tone lateralizedto the positionof the initial event.Various ICIs and but thatmonotonically increase with decreasingleft azimuthsin theright ongoingdurations were tested.In sucha context,the precedenceeffect, (i.e,, the distant) ear. Thesepredictions were, indeed,confirmed by re- the Franssenillusion, and the formationof auditoryobjects will be dis- sultsof maskingexperiments using the same stimuli presented monaural- cussed.[Work supportedby NSE and NIH.] ly in the two ears.[Work supportedby the Departmentof VeteransAf- fairs.] 11:00 D10. Minimum audible movementangles for sourcesvarying in both 11:30 elevation and azimuth. David R. Pertort and Kourosh Saberi (PsychoacousticsLaboratory, California State University, Los Angeles, D12. Effects of temporal positionon the detectability of interaurally uneorrelated noise. Martin E. Rickert and Donald E. Robinson CA 90032) (Department of Psychology,Indiana University, Bloomington,IN A minimumaudible movement angle (MAMA) experimentwas per- 47405 ) formed in a "simulatedmotion" paradigmwhich allowedtravel along vertical,horizontal, or obliquepaths. Velocities of travelranged from 1.8- Listeners'ability to detectan interaurallyuncorrelated segment of noise within binaural broadband noise bursts was measured as a function 230ø/s.With motion restrictedto the horizontal plane, MAMA thresh- oldsincreased from about 1.7ø at the lowestvelocity ( 1.8 ø/s) to roughly of the temporalposition of the uncorrelatednoise. Psychometric func- 10ø at a simulatedvelocity of 230ø/s. With the sound traveling on an tions were obtainedby varying the duration of uncorrelatednoise. The obliqueplane (45 ø rotation above the horizontal) MAMA thresholdsgen- resultsindicate that detectability improves as the duration of uncorrelated erally matchedthose of the horizontal condition.When the motion was noiseis increased.Temporal position affected detectability only at the restrictedto the verticalplane (the array wasrotated 90 ø from the hori- shortestdurations of uncorrelatednoise tested; a segmentpositioned ei- zontal) MAMA thresholdswere substantially higher at all velocitiesthan therat theend or at thebeginning of theburst is more detectable than one thresholdsobserved in the other array orientations,often exceeding8 ø in the middle.This resultis not easilyaccounted for by modelsof binaural evenat low velocities.Post hoc tests conducted at obliqueplanes 80 ø and processing.For example,the EC model[N. I. Durlach,J. Acoust.Soc. 87ø indicated that the slightestdeviation from the vertical,had a signifi- Am. 35, 1206-1218( 1963) ] assumesthat performanceis determinedby cant impacton MAMA thresholds,A deviationof 10ø from the vertical thepower of thedifference in thewaveforms in thetwo auditory channels. orientation was sufficient to reduce thresholds to those obtained when the Sincepower is computedover the total stimulusduration, it is indepen- soundmoves on the horizontalplane (0ø rotation abovehorizontal). dent of the temporalposition of the uncorrelatednoise. The relationship Theseresults suggest: ( 1 ) the abilityto detectmotion is essentiallyinde- of the current findingsto thosefor the discriminabilityof diotic noise pendentof the path traveledwith onenoted exception; sources traveling stimuliwill be discussed.[Work supportedby AFOSR.] on an absolutevertical dimension,and (2) detectionof the direction of travelis possiblewith azimuthchanges of only a few tenthsof a degreeif concurrentchanges in elevationare available.Some implications of these resultsare discussed.[Work supportedby NSF and NIH.] 11:45 D13. Binaural detectionwith reproducibleshort-duration narrow-band maskers. E. Colburn and H. Steven Colburn (Boston University, 48 11:15 CummingtonStreet, Boston, MA 02215) Hit and false alarm rates were measured for detection ofa 500-Hz tone Dll. Masking in simulatedfree-field. Pierre L. Divenyi (Speechand HearingResearch, V.A. MedicalCenter, Martinez, CA 94553) targetin eachoften samplesof l/3-oct noisecentered at 500 Hz with a duration of 10 ms, for both NoS•r and NoSo conditions.Two target wave- R. W. Gatehouse[Proc. 12th I.C.A. 2, B2-7 ( 1986); J. Acoust. Soc. formswere used, differing only in their phaserelative to the masker.Three Am. Suppl. I 82, S108 (1987) ] reporteda free-fieldmasking effect that subjectsare being tested,but extensivedata is availablefrom only one appearsanomalous on the surface:A pure-toneprobe positionedat a subjectso far. Also, the NoSo experimenthas been done for only one nonzero azimuth is maximally maskedby a broadband noise when the target phase.The NoSo hit ratescorrelate very well with the energiesin azimuthalposition of the maskerdoes not coincide with that of the probe. the waveforms.For NoS•r,the effect of targetphase on hit rateis substan- In the presentexperiments, similar resultswere obtainedwith stimuli tial, with almost no correlationbetween the rates for the two target presentedin a simulatedfree-field using head-related transfer functions phases.This is very differentfrom earlierresults with longduration wave- measuredin an artificial head [J. Blauert and P. Laws, Acustica29, 273- forms [e,g., Gilkey et al., J. Acoust.Soc. Am. 78, 1207-1219 (1985)]. 277 ( 1973) ], with a 3-kHz probelocated at a 45ø left azimuthand a white- Neithermodels based on lateralposition nor modelsbased on the sumof noisemasker positioned at either45 ø, 30 ø, or 20ø left azimuths.Maximum the squaresof the interauraldifferences correlate well with the data.Since maskingwas observed at the 30ø noise location. This nonmonotonicmask- subjectsreport multiple images and the interauraldifferences sometimes ing effect may not be due to anything but the head-induced diffraction change substantially over the 10-ms duration, models that process the patternscharacteristic to the 3-kHz range: Measurementof the noise first and secondhalves of the waveformsare analyzedseparately. [Work power (at the outputof the transferfunctions insidethe l/3-oct band suportedby NINCDS. ] S12 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S12 TUESDAY MORNING, 28 NOVEMBER 1989 DIRECTORS' ROW 43, 8:30 TO 11:55A.M. SessionE. StructuralAcoustics and Vibration I: StructureborneNoise in NonhomogeneousSolids Miguel C. Junger,Chairman Cambridge,4coustical Associates, Inc., 80 ShermanStreet, Cambridge, Massachusetts 02140 Chairman's IntroductionS:30 Invited Papers 8:35 El. Steady-stateand transient waves in random composites.Abraham I. Beltzer (Holon Institute for TechnologicalEducation affiliated with TeI-Aviv University, TeI-Aviv, Israel and Laboratorie de Mod61isationen M6eanique,Universit6 Pierre et Marie Curie,France} A causalapproach is applied to investigateharmonic waves in randommedia, which exhibit both scattering and viscoelastielosses. Since this approach allows for evaluationof the effectivewavenumber for the entire frequencyinterval, the transientsmay then be investigatedby the classicalFourier method.A particular exampledeals with thescattering model by Wu, andthe viscoelastic model by Azimi, whichfind applications in seismology.Numerical results for a transientresponse of this Wu-Azimi modelare presented.The pulse reinforcingand delayingeffects of multiplescattering are shownto be interrelatedthrough a causalitycon- staint. 9:00 E2. The constructionof dynsmlceffective medium theories (EMT} for inhomogeneousfluid or elastic continua.G. C. Gaunaurd(Naval SurfaceWarfare Center, Research Department (R43), WhiteOak, Silver Spring,MD 20903-5000) A deterministictechnique is describedthat predictsthe dynamic (i.e., frequency-dependent)effective propertiesof inhomogeneousmedia. Situations are consideredin whicha hostmedium sustaining a distribu- tion of sphericalinhomogeneities is either fluid or solid.In the Rayleighregion (viz., a/It < I } in whichmost effectivemedium theories (EMTs) hold,inhomogeneities of any shape can be assumed spherical without much error.However, assuming them to bepenetrable implies that theyrealistically admit interior fields, which are coupledto theexterior ones by stress and displacement boundary conditions. The outcome of dynamicEMTs is a setof expressionsto predictthe frequencydependence of the effective(sound) wave speed •cn, and the effective(sound) waveattenuation &cer in the mixture.These quantities emerge from the real and imaginary partsof the effective wavenumber kcer, which are connected byKramers-Kronig type relations, ashas been discussedby Dr. Beltzer.An importantconsideration for thepractical usefulness of an EMT is itssimplicity. Often,the resulting predictive expressions are so cumbersome that they can only be evaluated by means of quite formidablecomputer codes. The principalEMT to the discussedhere, yields closed-form simple predictive expressionsfor •:c•rand &eer,which, in manyinstances, are valid over significantlybroad frequency bands. Theseexpressions reduce to (and contain)almost all existingearlier results for static(f•0) cases.The (presentform of this) EMT is constructedby exploitingthe presenceof the monopoleresonance of the inhomogeneities,and thusit worksbest in the casesin whichthis resonanceis dominant.This EMT hasto be furtherrefined to accountfor caseswhere the dipoleresonance is dominant.In thisregard, Dr. Martinez,in a later lecture,will discussan approachto more accuratelydeal with heavy/rigidinclusions in a rubberlike matrix. [Work supportedby NSWC and ONR. ] 9:25 E3. Someacoustic properties of layered-compositeplates. Y. F. Hwang and P. J. Zoccola (David Taylor Research Center, Bethesda, MD 20084) This paper discussesa method of calculatingthe modesof wave propagationin a compositeplate. A compositeplate consists of layersof differentelastic and/or vlscoelasticmaterials. A novelthree-layer sand- wich plate, which utilizesa light weightsoft core sandwichedbetween stiffer septums (to increaseflexural rigidity without severeweight or cost penalty), will be closely examined in terms of its antisymmetrical (flexuralor bending)and symmetrical(longitudinal or bulge)modes of wavepropagation. The dispersion relationshipand the coincidencefrequency of the flexuralwave will bediscussed. Effects of fluid loadingand someother acoustic properties of thecomposite plate, such as the transmissibilityand reflectivity of sound,will also be discussed. S13 J.Acoust. Soc. Am. Suppl. 1, VoL 86, Fall 1989 118thMeeting: Acoustical Society of America S13 9:50 F_A.Pulse propagation in irregularlyheterogeneous waveguides--Some model calculations. John J. McCoy andJohn Tanzosh(School of Engineeringand Architecture,Catholic University of America,Washington, DC 20064) Numericalsimulations are presented of the propagationof pulsesthrough a seriesof idealizedone-dimen- sionalwaveguides, both isolated and coupled. The resultsare discussed in thecontext of determiningwhat aspectslead to predictivemodeling in termsof statisticalmoments; what ergodicities can be argued to allow predictionsfor an ensembleto applyto a singlemanifestation of the ensemble;what physicsneed to be accommodatedin meanfield formulations;what are the necessarystructures of theseformulations; etc. The emphasisof the simulationsis to illustratephenomena, as distinct from quantitativelyreproducing expected resultsof physicalexperiments for specificapplication. [Work supportedby ONR.] 10:15 E5. The dampingof beam vibrationsby viscoelasticlayers bearing pellets: Theory and experiment. Rudolph Martinez (CambridgeAcoustical Associates, Inc., Cambridge,MA 02140) A numberof powerfulsolutions for thescattering action of sphericalinclusions in extendedelastic media areavailable in theliterature. These analyses are relatively general and thus are usually too complex to provide directinsight into the relationsbetween their manydesign variables. Rather, they oftenrequire parametric studiesbased on numericalresults. The presentdevelopment takes instead a simplerview that focuseson the two only possibletypes of motionof a singleunderformable pellet in an incidentfield containing dilatational andshear waves: a translationalresponse for thedilatational case and both pellet translation and rotation for the shearwave field. The analysisemphasizes the fact that eachtype of motionhas an associatedbroad resonanceand antiresonance pair about which energy dissipation is maximized.These frequencies depend on pelletmass and moment of inertia,and may therefore be controlled to increasedamping within given bands of interest.Our studyof flexuralvibrations applies the single-pellettheory to predictthe effectiveloss factor of beamshaving constrained and unconstrained layers bearing undeformable inclusions, and compares the mod- el's results to measurements. ContributedPapers 10:40 The airbornetransmission loss across resilient mounting systems is investigatedby developingan analyticmodel of the transmissionof vibra- E6. Transientpropagation of a one-dimensionalwave in a harmonically tion froma point-driven,simply supported plate to a parallelinfinite plate inhomogeneousmedium. Hyun Sil Kim andJerry H. Ginsberg(School throughthe air separatingthe plates. Predictions of airbornetransmission of MechanicalEngineering, Georgia Institute of Technology,Atlanta, lossesare comparedwith transmissionlosses through resilient mounts to GA 30332) quantifythe roleof the airbornepath in transmissionlosses through well- This paperconsiders the transientpropagation of a one-dimensional designed,single-stage resilient mounting systems. The sensitivityof the dilatationalwave in a spatiallyharmonic inhomogeneous medium of airbornetransmission losses to separationdistance, thickness of the re- semi-infiniteextent. A regularperturbation analysis of theresponse, suit- ceivingplate, size of the drivenplate, and dampingin the platesare pre- sented. ablewhen the amplitude of the inhomogeneousdeviation is small, reveals that inhomogeneitycauses backscattering in the second-orderterm. A temporallyharmonic excitation generating a signalwhose wavelength is twice the periodicityof the inhomogeneityis shownto leadto a second- orderstanding wave behind the wave front whose amplitude growth ulti- 11:10 mately invalidatesthe solution.The frequencyat which this situation occursmatches the frequency at whichFloquet theory predicts instability ES. Embeddedfiber-optic acoustic transducerand sensorarrays for in the steady-stateresponse. A finite-differenceanalysis based on the acousto-ultrasonicevaluation of materials. Chris Thompson and methodof characteristicsconfirms the early-time validity of theperturba- Richard O. Claus (Fiber and Electro-OpticsResearch Center, Virginia tion solution.It alsoreveals that the late-timesignal is essentiallya stand- PolytechnicInstitute and StateUniversity, Blacksburg, VA 24061) ing wavein the regionextending outward from the transducerto within severalwavelengths of thewave front, whereas it mostclosely resembles a A noveltechnique for generatingacoustic pulses inside a materialand nondispersivepropagating wave in the region of the wave front. [Work then sensingthe signalagain within the material is proposed.A hollow supportedby NSF. ] core fiber-opticwaveguide is constructedwith periodicsolid "plugs" fusedin place.The opticalfiber is embedded in a solidmaterial, and laser light is sentdown the hollowcore fiber. The laserenergy is absorbedand attenuatedat the plug locations,causing localized heating that will then act as a source of ultrasonic waves due to the thermoelastic mechanism. Fiber-opticsensors are also embedded to receivethe acoustic wave fronts. 10:55 Any materialdisturbances in the path of the acousticwaves will effectthe velocityand attenuationof the wave,which can be quantifiedusing the E7. Airborne sound transmission across resilient mounts. Michael stresswave factor (Vary et al.). The fiber-optic transducer/receiving F. Shaw (Naval Sea SystemsCommand, Code 55N, Washington,DC pairscan be locatedanywhere within or upona materialor structureand 20375) and Courthey B. Burroughs (Applied ResearchLaboratory, can be usedto detectdamage in situ.This techniqueis particularlyvalu- PennsylvaniaState University, State College, PA 16801) ablein assessingdamage in compositematerials. S14 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S14 11:25 integraltheorem to evaluatenear- and farfieldpressures. Results include radiation patternsshowing the influenceof stiflenet size and location. E9. Analysis by variational principlesof the vibration and acoustic { Work supportedby ONR. I radiationof a vibratingelastic plate with circumferential stiflenets. Jerry H. Ginsbergand James McDaniel (School of MechanicalEngineering, GeorgiaInstitute of Technology,Atlanta, GA 30332) 11:40 Thisstudy concerns the effects of circumferential ring stiflenets on the El0. Noncompressibleliquid layer elasticity phenomenon.Vladimir vibrationand acoustic radiation of a submergedcircular elastic plate. The A. Babeshko(Department of Mathematical Modelling, Kuban State disk,which is excitedby an arbitraryaxisymmetric harmonic force, may University,Krasnodar, 350640, USSR) besupported by either a rigidannulus or by an infinitebaffle. The analysis employsvariational principles [J. H. Ginsbergand P. T. Chen,J. Acoust. A vibrationprocess localization phenomenon, arising in semi-infinite So<:.Am. Suppl.1 82, SI (1987) ], in whichsurface pressure and velocity regions,is knownto existin mediaother than solid deformable media. An are representedby assumedmodel functions. The presenceof stiflenets infinitelayer of an idealincompressibleliquid r + h deepis in contactwith altersthe kinetic and potential energies of theplate, but it doesnot affect an elasticplate, vibratingat the frequencyto. Analysisof this system the formulation of lbe fluid-structure interaction. Two structural formu- showsthat for somevibrating plate size and mass,energy propagates to lationsare addressed,differing by whetherthe connectionbetween the infinity unattenuated.The conditionsfavoring such vibration processes plateand the stiflehersis implicitlyor explicitlyenforced. The resultsfor when unrestrictedresonances can ariseare presented[ V. A. Babeshko, surfaceresponse are usedin conjunctionwith the Kirchhoff-Helmholtz Lett. J. Technol. Phys.14 (8} (1988)I- TUESDAY MORNING, 28 NOVEMBER 1989 DIRECTORS' ROW 46, 9:00 A.M. TO 12:05 P.M. SessionF. Musical AcousticsI: Strings, Winds, and Harmony ThomasD. Rossing,Cochairman Departmentof Physics,Northern Illinois University,DeKalb, Illinois 60115 Uwe Hansen, Cochairman Departmentof Physics,Indiana State University,Terre Haute, Indiana 47809 Chairman's Introductions9:00 Contributed Papers 9:05 Following a suggestionof Benade (private communication)that smallinharmonicities in the mode frequencies of a stringmay have impor- Fl. Determination of a string's impulseand step responsesby Laplace tanteffects on its bowingresponse, a method has been developed for pre- transformation;relationshi• to exciting mechanisms.Ernst Terhardt ciselymeasuring the (complex)normal frequencies and polarizations of a (Institutefor Electroacoustics,Technical University, D-8000 Munich, stringas mounted on a violin,with the aim of determining,among other WestGermany) things,whether the violin maker'sadjustments orbridge and soundpost The excitationof a musicalstring is intimatelydependent on the are systematicallyreflected in the string'sinharmonicity. The procedure, string'sdynamic behavior--i.e., of thestring alone--at the pointof excita- whichuses acoustic excitation and a two-dimensionalphotoelectric detec- tion. With resortableapproximation that behaviorcan essentiallybe re- tor of stringmotion, is basedon earlierwork [E. B. Arnold andG. Wein- gardedas that of a linear system.An elegantway to describeit is by reich,J. Acoust.Soc. Am. 72, 1739(1982) ] but usesa higherfrequency Laplacetransformation. On the basisof the familiarequivalence of string resolution(0.125 Hz) and a least-squaresfitting of the resonantstring andhomogeneous electrical transmission line, the string impedance at the response.Some results on theshift of stringfrequencies due to controlled excitationpoint is expressed in Laplace-transformnotation. This serves as loadingof thebridge will bepresented. [Work supportedby NSF.] a basisfor gettingthe frequency characteristics of theimpedance (magni- tudeand phase), and, by inverseLaplace transformation, the impulse and stepresponses of forceand velocity,respectively. These characteristics of the stringalone prove to be very helpful in discussingthe interaction 9:35 betweena stringand any excitingdevice such as a bow or pianohammer. F3. Low pitchednotes associated with large bowingforces. R. J. Hanson, A. J. Schneider(Physics Department, University of Northern Iowa, Cedar Falls, IA 50614-0150),and F. W. Halgedahl(School of Music, Universityof Northern Iowa, CedarFalls, IA 50614) 9:20 For bowingforces greater than the Schelling maximum [J. Shelling,J. F2. Preeislon measurementof violin string modes. Bernard Ricca, Acoust.Soc. Am. 53, 2(•41 ( 1973}] there are variousperiodic motions Gabriel Weinreich,and NicholasR. Michael (Randall Laboratoryof possiblewith periodsmuch longer than any naturalperiod of the string Physics,University of Michigan, Ann Arbor, MI 48109) [M. E. Mcintyre, R. T. Schumacher,and 1. Woodhouse,I. Acoust.So:. S15 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S15 Am. 74, 1325-1345(1983)]. Aperiodicmotion with a corresponding stick velocityas it crossesthe trigger plane.The algorithmgives very raucoussound are, however,much more readily produced with a large reliabletriggers over a biggerdynamic range than existingsynthesizer bowingforce. If careis exercised by theplayer, a repetitivewaveform with keyboards. a nonraucoussound for verylarge bowing forces can be maintained for as longas several seconds. The resultingpitch that is heardis in agreement with the fundamentalfrequency of the harmonicseries obtained with a frequencyanalyzer. This is not the small flattening effect with moderately 10:20 highbowing force analyzed by Mcintyre et al. butis a largeeffect produc- ing pitchestypically ranging from a musical6th to a musical12th below F6. Chaoticdynamics of woodwindmultiphonics. Douglas H. Keefeand the normalpitch. Systematicpatterns of the measuredfrequencies as a Bernice Laden(Systematic Musicology Program, School of Music,DN- functionof normal stringfrequencies have beenobserved and possible 10, Universityof Washington,Seattle, WA 98195) interpretationswill be presented. Experimentalevidence for chaosin multiphonictones derives from measuredpower spectra, and more directlyfrom the measuredcorrela- tion dimensionof the reconstructedphase space of the dynamicalsystem. The frequencies•.m of the line spectralcomponents of a reed-driven woodwindmultiphonic fit a biperiodicspectrum of low- to mid-playing levels;i.e., thereexist base frequenciesf• 10:35 10:05 F7. Effect of pressurerise time on starting transientsof flue pipes. A. FS. Triggering•lgorithmfor the Boie radiodrum. Max Mathews (Music W. Nolle (Department of Physics,University of Texas, Austin, TX Department,Stanford University, Stanford, CA 94305-4060) and R. 78712) and T. L. Finch (Department of Physics,St. Lawrence A. Bole (AT&T Bell Laboratories,Murray Hill, NJ 07974) University,Canton, NY 13617) The Boieradio drum is used to playelectronic music synthesizers. The A specialwind supplyprovides adjustable rise time T,. Adjustable drum is a sensitivesurface that tracksthe positionand velocityof two or closedand openpipes having frequencies 198 to 256 Hz are studied.The more drum sticks in a shallow volume above the surface in three dimen- steady-statesupply pressure is 530Pa. The precursor,at severaltimes the sions.Positions and velocitiesare determinedby measuringthe strength fundamentalfrequency, and a subsequentburst dominatedby the next and distributionof electricalcapacitance between the tips of the drum modeabove the fundamental are strongest for Tr between1 and 10funda- sticks and the surface. The drum is made force sensitive in the surface mentalperiods, but may vanishfor largeror smallerTr. For nearlyhar- normaldirection by coatingeither the sticksor surfacewith an elastic monicpipe resonances, the fundamentaland highercomponents can grow foamlayer. Position information of a stickin contactthen corresponds to simultaneously,and the burst may exceed the steady-state amplitude. For two-dimensionalposition on the surface and force into the surface. One of more inharmonicmodes (e.g., a wider pipe) the fundamentalquenches the drums functionsis to generatetriggers that start notes.The trigger the burst when the overallamplitude is well belowsteady state. External algorithmmust be very reliablein that one and only one triggeris pro- soundat the fundamentalfrequency suppresses both the precursorand ducedno matter how gentlyor how hard the drum is struck.An algo- thesubsequent burst. Important features of theexperimental results occur rithm, patternedafter the actionof a pianokey hasbeen implemented. A in a computermodel that involvestime-delayed nonlinear feedback and triggerplane and a resetplane are defined above the drum surface. Motion usessimple oscillators to representresonator modes [N.H. Fletcher, through the trigger plane producesa trigger. Motion back through the Acustica 34, 224-233 (1976)]. With numerical rather than analytical higherreset plane enables the next trigger.Loudness is controlledby the solutionsit is possibleto treat the realisticcase of the fully switchedjet. 10:50-11:05 Break S16 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S16 11:05 Thedevelopment of culturally specific perceptual knowledge for mu- FS. Perceptionof a musicalconstruct by professionalmusicians. sicwas investigated. Using a background/contrastparadigm originally PhilipBaczewski and RosemaryN. Killam (ComputingCenter, developedfor speechperception [Eilers et al., J. SpeechHear. Res.20, Universityof North Texas,P.O. Box 13495,Denton, TX 76203) 766-780 (1977)], Western10-13 year oldsand adultswere tested in detectionof mistuningsin melodies based on native Western major and Thisstudy examined 15 highly trained listeners' response toa stimulus Westernminor scales and a non-nativeJavanese pelog scale. Each age consistingof a 16-measurepiece of music.Subjects were music perfor- groupwas divided equally into musically inexperienced and musical sub- mancegraduate teaching fellows. They were presented with a tapedper- jects.The musicallyinexperienced adults had a perceptualbias for the formance(prepared by professionalmusicians) of threerepetitions of a melodiesbased on their native scales,but the adult musiciansdid not Mozartduet, together with theexperiment's instructions. Experimental appearto sharethis bias, since they were able to generalizetheir knowl- designallowed measurement of response to each stimulus repetition. Sub- edgeof Westernscales to the perception of a melodybased on the Javanese jectresponses were recorded in oneof twodiffering formats: eight subjects scale.In contrastto the adults, the musician 10-13 year olds had a percep- recordedthe musical notation of eachof thethree stimulus repetitions in a tual biasfor the melodiesbased on Westernscales, and the nonmusician differentcolor of pen;seven subjects recorded the musicalnotation of each 10-13 yearolds did not sharethis bias.The musicallyinexperienced of the threestimulus repetitions on a differenttranslucent paper overlay. adults'performance was significantly better than that of the musically Subjectswere allowed 3 min to recordeach response. Data wereencoded inexperiencedchildren on boththe Westernand Javanese scales, and the and analyzedto providemeasurement of cumulativeresponse arising samewas true for themusician adults and children, indicating develop- from all three repetitions.Effects of pitch boundaryand constituent mentof perceputalability that is independent of musical training. Percep- groupingswere obtained. A modelemerges of complexperceptual strate- tionof Westernmistunings was significantly related to thechildrens' total giesemployed by musically sophisticated subjects. The strategies incorpo- yearsof musicalexperience and the amount of timedevoted to practicing rate pitch relationships,contour, and grouping.[Work supported,in a musicalinstrument or voice.[Work supportedby theAmerican Psy- part, by an Apple EducationFoundation Grant.] chologicalAssociation. ] 11:20 F9. Vocalrange for speechdetermines perception of a musicalpattern. Diana Deutsch,Tom North,andLee Ray (Departmentof Psychology, Universityof California,San Diego, La Jolla,CA 92093) 11:50 Previously,a two-tonemusical pattern was described that possesses Fll. The developmentof musical tuning perception from infancy to somehighly unexpectedproperties [D. Deutsch,J. Acoust.Soc. Am. adulthood.Michael P. Lynch, RebeccaE. Eilers, D. Kimbrough Oiler Suppl.1 80, S93 (1986) ]. For anygiven listener, it isheard as ascending (Departmentsof Psychologyand Pediatrics, Mailman Center for Child whenplayed in onekey, and yet as descending when played in a different Development,University of Miami, P.O. Box 016820, Miami, FL 33101), key.As a furtherparadox, when the pattern is played in anykey it isheard andPaul Wilson(DepartmentofMusicTheory/Composition, Schoolof asascending by some listeners but as descending by others.Since then, the Music, Universityof Miami, Coral Gables,FL 33143) behaviorof this musicalparadox has been explored under parametric variation[e.g., D. Deutsch,Percept. Psychophys. 42, 563-575 (1987)]; however,the basisfor the individualdifferences remained elusive. Here, it Three prominenthypotheses were evaluatedby testingWestern in- is demonstratedthat a strikingcorrelate exists between the way this pat- fants'and adults' abilities to detectmistunings in melodiesbased on native ternis perceivedand position of thelistener's vocal range for speech.To Westernand non-nativeJavanese scales. The "innatehesshypothesis" of the authors'knowledge, this is the firstdemonstration of a closeconnec- speechperception applied to musicperception suggests that infantsare tionbetween the perceptionof a musicalpattern on theone hand and the bornwith an equipotentialto perceivethe scalesof any culture.The "nat- listener'sspeech characteristics on theother. Implications of thesefind- uralnesshypothesis" of auditoryperception suggests that the ear is best ingsfor bothmusic and speech are discussed. [Work supportedby the suitedfor processingthe small-integerfrequency ratios that comprise Digital EquipmentCorporation, and by the UCSD BiomedicalResearch Westernscales. Finally, the "perceptualloss hypothesis" of speechper- Fund. ] ceptionallied to musicperception proposes that the abilityto perceive non-nativescales is reducedand the ability to perceivenative scalesis maintainedfrom infancyto adulthood.In supportof the innatehesshy- pothesisand in contradictionof the naturalnesshypothesis, 5- to 7- 11:35 month-oldinfants did not possessa perceptual bias for eitherWestern or Javanesescales. The perceptualloss hypothesis was weakenedbecause, F10. Native and non-native music perception by children. Michael althoughadults possessed a perceptual bias for the Westernover the Java- P. Lynch (Departmentsof Pediatricsand Psychology,Mailman Center nesescales, adults' perception of both thesescale types was better than for Child Development,University of Miami, P.O. Box 016820,Miami, thatof infants.[Work supportedby theAmerican Psychological Associ- FL 33101) ation. ] S17 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $17 TUESDAY MORNING, 28 NOVEMBER 1989 ST. LOUIS BALLROOM A, 9:00 TO 11:00 A.M. SessionG. SpeechCommunication I: Linguisticsand Phonetics Patti J. Price, Chairman SRI International,EK178, Menlo Park, California 94025 ContributedPapers 9:00 The Kabardianvowel system has been the focusof muchdebate. Cen- tral to this debateis whetheror not the low vowelo is quantitatively G1. Prosody as a cue to syntactlc structure:The role of prosodic distinctfrom the mid vowel a. Thisstudy examines phonetic evidence that informationin disamblguafingphonetically ambiguous sentences. Patti bearson thisissue. Spectrographic and durational analyses were conduct- J. Price (SRI International, EK 178, Menlo Park, CA 94025), edfor 118vowels taken from a Kabardiantext readtwice, once slowly and Marl Ostendorf (Boston University, Boston, MA 02215), onceat a normal rate, by three adult male speakers.Each vowel was StefanieShattuck-Hufnagel (MIT, Cambridge, MA 02139), and measuredfor its firstthree formant frequencies and duration.The data Cynthia Fong (BostonUniversity, Boston, MA 02215) revealthat the meanduration of • is 62.5% that of a, comparedto the To gobeyond dictation tasks, speech recognition needs to beintegrat- meanduration ofi whichis 67.5% thatof•. Thesefigures represent ratios ed withan understandingcomponent. Natural language understanding characteristic of intrinsic durational differences rather than distinctive research,however, has typically considered text input, and has not had to lengthand supporta three-vowelanalysis. Qualitatively, the formantfre- facethe massive phonetic ambiguity arising from speech input: Even with quencymeasurements support the claims in the literaturethat Kabardian no phonetalcerrors, word boundary ambiguities can yield thousands of vowelscontrast uniquely along the heightparameter with somefront- parsesfor a substantialportion of moderatelength sentences. This study backallophony for the two highervowels. Moreover, u hasa meanF 1 of assessesthe potentialrole of prosodyin aidingspeech understanding. 690 Hz as comparedto • which exhibitsa meanF 1 of 510 Hz, further Usingphonetically similar sentence pairs with andwithout a majorstruc- supportinga three-vowelanalysis. [Work supportedby NSF.] turalbreak, seven types of structuralambiguity were investigated. Profes- sionalspeakers read ten sentencesof eachtype after reading diambiguat- ing contexts.Naive listenersselected the appropriatecontext for each sentence.For recordingand listening, only one member of eachpair was 9:36 used in each session;sentences were otherwiserandomized. The results G4. Phonotacticconstraints as a filterof thearticulatory-acoustic signal. indicatesubstantial differences across structural types: e.g., appositions Maria-JosepSo16 (Laboratori de Fon•tica, Facultat de Lletres, and tagswere nearly always correctly identified, whereas main-main ver- UniversitatAut6noma de Barcelona, 08193 Bellaterra, Barcelona, Spain) susmain-subordinate pairs were at the chancelevel. This suggeststhat prosodycan disambiguatesome, but not all, structuralambiguities. Thepresent study investigates the working and status of phonotactic [Work supportedby NSF.] constraintsin themodeling of speechproduction and speech perception. Phonotacticconstraints act as a filterof theacoustic signal and the subse- quentarticulatory gestures. The adaptationof foreignborrowings and foreignlanguage learning provide many examples of the transferof the phonotacticconstraints ofL 1to the perception and production of L 2, e.g., 9:12 Englishspeakers' pronunciation ["nou"sei] for the Spanishsequence G2. Intonation and syntax: Preferred co-occurrencesbetween pitch ["no"se]"Idon't know," and [el ei "tret"bell for theFrench sequence contour and sentence structure. Aart de Zitter and Ren6 Collier [el e "tre"bell "sheis verybeautiful," shows how the English syllable (Institutefor PerceptionResearch, P.O. Box 513, 5600MB Eindhoven, structureconstraint that prevents short vowels from occurring in stressed The Netherlands) opensyllables is imposed on theselanguages. When this syllable structure constraintis not present,though, as in the words [ell "elle," [bell Melodic models,also called "intonation grammars," specify the per- "belle,"English speakers pronounce a short vowel.To checkwhether ceptuallyrelevant properties of pitch movementsand indicatewhich se- higher-orderphonotactic constraints act as a filterof theacoustic signal, quencesof themconstitute well-formed pitch contours IN. Willems,R. the samephysical signals were placed in differentsyllabic structures by Collier, and Y. 't Hart, J. Acoust. Soc. Am. 84, 1250-1261 (1988)]. It digitalmanipulation with an ILS program.The stimuli were presented to appearsthat for eachbasic intonation pattern several contour variants can speakersof differentlanguages to checkwhether they perceived the same begenerated that sound equally acceptable from a purelymelodic point of stimulidifferently depending on the phonotacticconstraints of theirlan- view. However,not everycontour is compatiblewith just any sentence. guage.The resultsshow that phoneticand phonological identification is Indeed,in manycases the choicesthat speakersmake appear to be far overriddenby phonotacticconstraints, which advocates for a top-down fromrandom. Rather, they are governed by syntactic, semantic, and prag- processingof the signal. matic considerations.The presentpaper concentrates on syntacticvari- ablessuch as the degreeof structuralcohesion between pitch-accented words or the depth of the syntacticboundary betweenphrasal constitu- ents.Experiments will bediscussed in whichsyntactic variables appear to influencethe preferenceof the speakeror the listenerfor particularpitch 9:48 contours. It will be examined to what extent these results are relevant for improvingsynthetic intonation in text-to-speechsystems. G$. Finding and usingstress accent in spokenlanguage understanding. ShirleyA. Steeleand Suzanne A. Liebowitz(Unisys Corporation, P.O. Box 517, Paoli, PA 19301 ) 9:24 Stressaccent is a featureof the prosodicsystem in Englishthat func- tions syntagmaticallyto organize an utterance into shorter sections. G3. Phonetic evidence for a three-vowel system in Kabardian. John- Knowledgeabout the locationof stressaccents in the segmentalstream DongwookChoi (Departmentof Linguistics,UCLA, LosAngeles, CA can help a speaker-independentspoken language understanding system 90024-1543 ) prune the searchspace, e.g., for the correctlexical item or the correct SlS J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S18 pragmaticinterpretation. Unlike automaticsystems, human listeners are thesetwo modules,for example,by creatinga file containingthe fully ableto locateaccents because they know what accentedversus unaccent- specifiedfeature descriptions of thesegments in a groupof languages,or ed syllablessound like and because they can normalize for suchthings as all segmentsdefined by a selectedset of featurevalues. Such files can be speakerand speaking rate. Here, the resultsof a studydesigned to locate exportedto a standardstatistics package for sophisticated processing, but accentsin continuousspeech are reported. An attemptis made to useboth simplercounting operations can be performedwithin the UPSID pro- speechknowledge and normalization by identifyingsalient characteristics gram. of accentin trainingdata along with accentcharacteristics of the test utterance.Accented and unaccented syllables are separated using selected clusteringalgorithms on a featurespace of energyand duration measures. In addition,some specific ways in which the outputof the accentfinder can be usedby othercomponents of the system(e.g., lexicalaccess, prag- 10:24 matics)are suggested.[Work partiallysupported by DARPA/ISTO.] G8. Accentedhessand givenness:Change in word, syllable,or segment duration? Wieke Eefting and Sieb G. Nooteboom (Institute of Phonetics,University of Utrecht, Trans 10, 3512 JK Utrecht, The Netherlands } In an earlierpaper [W. Eelting,J. Acoust.Soc. Am. Suppl.I 85, S97 10:00 (1989) ] theresults of a studyconcerning the relation between the features G6. More on domain-final lengthening and foot-level shortening in "givenness"and "accentedhess"and the durationof Dutch wordswere spoken English. Brad Rakerd, William Sennett, and Paul Cooke presented.Production data from a highly trained,professional speaker (Department of Audiology and Speech Sciences,Michigan State indicatedthat the duration of one syllableand three syllablewords is University,East Lansing, MI 48824) lengthenedwhen accented. Information value had only an effecton one syllable words: One syllable words containing new information were The stressedsyllables of spokenEnglish tend to be lengthenedwhen lengthenedas compared to the samewords containing old information.It theyoccur before major syntactic boundaries (domain-finallengthening) remained unclear whether the differences in word duration affected the andto beshortened when they are followedby unstressedsyllables within temporalstructure of the whole word; a local changewithin the word the samemetrical foot (foot-leoelshortening). The natureof the interac- might be responsiblefor the differences.In the presentstudy, the dura- tion betweenthese two factorshas beeninvesligated. A first study lB. tionsof the syllablesand the segmentswithin the targetwords are mea- Rakerd, W. Sennett, and C. A. Fowler, Phonetica 44, 147-155 (1987)] sured.Data will bepresented showing that all segmentsin theone syllable suggestedthat their effectsare independent.In that study,the syllabic wordsand all syllablesin the threesyllable words contribute to thediffer- structureof metricalfeet spanning word boundaries were manipulated, encesin word durationas describedabove. These findings support our wherethose boundaries either coincidedwith noun-phrase/verb-phrase claimthat in speechthe control of wordduration is an active skill and that boundariesor did not, and we comparedstressed syllable durations. A wordsare the mostimportant units of speechcommunication. [Work limitationon the comparisonwas that the phoneticdetails of the metrical supportedby the SPIN-ASSP Program.] feetof interestwere matched but not heldconstant across syntactic condi- tions. In the presentstudy, the procedureto control phoneticdetails throughouthas been refined. With thismodification, a significantinterac- tion was observedbetween domain-final lengtheningand foot-level shortening;specifically, it wasfound that the intervalsbetween lengthen- 10:36 ingsdelimited the domainsover which substantialshortening could oc- cur. The whole patternof resultswill be discussedin light of current G9. The syllabic attachmentof post-vocalicliquids and nasals:Error theoriesof foot-levelshortening. ellcitationevidence. Ann Stuart Laubstein(Department of Linguistics, Carleton University, Ottawa K I S 5B6, Canada) Inferencesbased on spontaneousspeech error corporaregarding the syllabicattachment of post-vocalicliquids and nasalsare experimentally testedusing the error elicitationtechnique described in Motley, Baars, and Camden (1983). In the first experiment,44 subjectswere testedon 10:12 "Vr" and "Vl" sequences.As predictedVr sequenceswere not split up, whereasVI sequenceswere. Subjects moved the vowelsignificantly more G7. Ul•lating UPSID. Ian Maddiesonand g.ristin Precoda(Phonetics Laboratory,Department of Linguistics,University of California,Los oftenin VI sequencesthan in Vr sequences(Wilcox0n, p <0.001); in Angeles,CA 90024) addition'T' wasmoved significantly more often than "r" (Wilcoxon,p <0.O301). Here, VI did, however,move as a unit, lendingsupport to the UPSID--the UCLA phonologicalsegment inventory database•is a Halle andVergnaud (1980) sonorantconstituent. The secondexperiment databasecontaining the phonemeinventories of a largegenetically based used45 subjectsto testthe viabilityofsuch a constituentby comparingVI sampleof languages[I. Maddicson,Patterns of Sounds(1984) ]. Each to vowelnasal ("Vn") sequences,and nasaland liquidobstruent (SO) phonemeis specified in termsof a comprehensiveset of phoneticfeatures. sequencesto obstmentobstruent (OO) sequences.As predictedVI and The firstversion of thedatabase has proven useful to scholarsinterested in Vn sequenceswere indistinguishablefrom eachother and SO and OO phonologicaluniversals and theoriesconcerning the structureof phono- weresignificantly different. This convergenceof evidencefrom separate logicalsystems le.g., B. Lindblomand I. Maddieson,in Language,Speech domainssupports the generatireassumption that constructsof compe- & Mind, editedby L. M. Hyman and C. N. Li ( 1988); K. Stevensand S. J. tenceunderlie the capacityfor use. Keyset,Language 65, 81-106 (1989) ]. An expandedand correctedsec- ondversion is currently in preparation.This versionimproves the sample, increasingcoverage of previouslyundersampled language families and correctinga fewoversampling errors, and correctingerrors in individual language inventories. A new custom-written software package for 10:48 MS-DOS systemsprovides economical and flexible means of storingand modifyingthis enhanced database and outputtingsubsets of the data for GI0. Classificationof Japanesesyllables including the speechsound furtheranalysis. The databaseis stored as several separate but interrelated found in loan words. Shizuo Hiki (School of Human Sciences.Waseda modulcs. One contains a listing of character codes for each distinct seg- University,Mikajima 2, Tokorozawa,359 Japan) ment type occurringin the databasepaired with a standardphonetic de- scriptionand with the list of featuresassigned to that segment.Another Methodsofclassifying Japanese syllables are discussed with regardto containsthe phonemeinventories as a set of charactercodes for each their usein evaluatingperformance of speechprocessing techniques for language.The databaseis principallyused by matinginformation from bothsynthesis/recognition systems and user's speaking/listening charac- S19 J. Acoust.Soc. Am. Suppl.1, VoL 86, Fall 1989 118th Meeting:Acoustical Society of America S19 teristics.First, by takinginto accountthe initial/intervocaliccontrast of plementingmissing kind of combinationsof thepreceding consonant and utterance, 16 intervocalicsyllables are introducedto the traditional 50 followingvowel, more than 60 newsyllables which can be pronounced in Japanesesyllables chart or the list of 100Japanese syllables that hasbeen the loan wordsare derived.Also discussedin this paper is rationalefor usedin articulationtest of speechtransmission channels. Then, by rear- variouslevels of derivingnew syllablesthat correspondto the special rangingthose syllables based on place/mannerof articulationand by sup- Kana lettersused in writingloan words in Japanese. TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM F, 1:00 TO 4:00 P.M. Session H. Engineering Acoustics I: Transducers, Materials, and Diffraction SungH. Ko, Cochairman Naval UnderwaterSystems Center, New London, Connecticut06320 StephenC. Thompson,Cochairman WestinghouseOceanic Division, Department 721, Cleveland,Ohio 44117 ContributedPapers 1:00 design,such as the numberof elementsversus resolution, be made.Field patternsproduced by a circular phasedarray, requiredto evaluatesuch H1. Tapered-thickness piezoelectric ceramics: Models and designtrade-offs, have not been extensively examined. In thispaper, a new characteristics. Peter G. Barth6 and Paul J. Benkeser (School of field calculationmethod, which requiresless computational time than ElectricalEngineering, Georgia Institute of Technology,Atlanta, GA existingmethods for a givenaccuracy, is presented.The curvedsurface of 30332) the array,radiating either continuous-wave or pulsedsignals, is subdivid- Tapered-thicknesspiezoelectrics have been studied for use in novel ed into incrementalareas which are small enoughthat the Fraunhofer devicesand imagingtechniques, such as widebandtransducers and fre- approximationcan be applied.The field pressureis computedby sum- quency-controlledbeam translation.However, no adequatetheoretical mingthe contributions from all theincremental areas using the Rayleigh- modelexists. In this work, a staircasemodel is developedin which the Sommerfelddiffraction integral. Theoretical resultsobtained with this thicknessgradient is representedas discrete steps. Each step is modeled as methodare comparedto thoseobtained using other methodsas well as a constant-thicknesssection, which contributes to the net response.This experimentaldata. The effectsof somedesign parameters such as the approachallows the resultsand techniquesdeveloped for conventional geometryof the array,pulse bandwidth, and amplitude weighting of the devicesto bereadily applied to thosewith taperedthickness; it emphasizes excitationsignal will alsobe presented. importantsimilarities and differences,and providesphysical insight into the dynamicsof taperedpiezoelectrics. The modelshows that the distrib- utedhalf-wavelength resonances dominate its response. In addition,how- ever,a taper-dependent,transverse interaction exists among steps, which 1:30 adds to the primary response.A coupledoscillator approach is usedto H3. Computation of the performance of an electret microphone model this transverse cross talk with an acoustic transmission line. The diaphragm. Peter L. Madaffari (Industrial ResearchProducts, 321 Bond theoreticalelectric input impedance,spectral content, pulse, and contin- Street, Elk Grove, IL 60007) uous-wavepressure profile responses obtained with the modelare shown to be in closeagreement with the measuredresponses from severaltapered An attempt was made to simulatethe performanceof a tensioned- piezoelectrictransducers. diaphragm,electret-based, microphone. Use of commercialfinite-ele- ment analysisprograms were unsuccessfulbecause of the computational systemused. An alternatemethod was developed using a relaxationap- proach. A third-order equationfor the balanceof forcesappeared to give approximatesolutions for both tensionedand multiply supported 1:15 diaphragms.Contours of physicaldeformation and acousticcompliance H2. Soundfield calculationfor a circular array transducer.Chankil Lee, weregenerated and mapped.From thesecomputations, an improvedelec- Intaek Kim, and Paul J. Benkeser(School of ElectricalEngineering, tret backplateand diaphragmsystem were designed. GeorgiaInstitute of Technology,Atlanta, GA 30332) Ultrasoniccircular phased arrays have long been employed in sonar, 1:45 NDT, and AGV applications.More recently,these arrays have been used in medicalimaging applications. The abilityof thesearrays to imagea 360ø H4. Edge effects in standard shapes piezoelectric transducers•A new field makesthem idealfor intravascularimaging applications. However, method for calculating elastic coefficients. Dov Hazony and intravascularapplications often require the diameterof the arrayto beon Birol Ozkaya (Department of Electrical Engineeringand Applied theorder of 1 min. The smallsize of the arraydictates that trade-offsin its Physics,Case Western Reserve University, Cleveland, OH 44106) S20 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S20 An electrical network similar to the Mason model is constructedto in a listeningroom. A knowledgeof self-and mutualradiation resistances conformwith the electroelastic properties of thinisotropic piezoelectric isalso important for an activesound power or noisecontrol. As an exam- ceramicsof standardshapes under the condition of shortstress pulses in ple, self- and mutual radiationresistances of two closelylocated loud- thisthickness mode. It will be seenthat thesepulses travel close to the speakersmeasured in a lectureroom are presented. Due to reflections,the speedof sound in anunbounded medium and that it ispossible to calculate resistancecurves measured in a roomare much more irregular compared the elasticcoefficients by settingthe two activefaces free. Specifically, to the resistancecurve of a pistonin an infinitebaffle. The effectof the bothA and#, theLam6 constants, may be calculated from one measure- loudspeakerplacement on theself and mutual resistances is very signifi- ment. cant. 2:45 2:00 HS. Bandwidthand absorptioneffects on intensificationin a structural- H5. A high-resolution beamformer for acoustic imaging. Donald C. Knudsen(Knudsen EngineeringLimited, Perth, Ontario, Canada) aenustieenelosure.LindaF. PerettiandEarlH. Dowell(Departmentof M•chanicalEngineering, Duke University,Durham, NC 27706) This paperdescribes the developmentof a high-resolutionimaging sonarsystem based on innovativedigital beamformingtechnology. The Asymptoticmodal analysis (AMA) providesa formal methodto ana- newbeamformer provides rapid access to a high-resolution,three-dimen- lyze high modaldensity sound fields in enclosures.Typically, uniform sionalimage field, scanningsimultaneously in both azimuth and eleva- responselevels occur throughout the interior. Exceptionsoccur at phys- ical boundariesof the enclosure, and at theoretical boundaries that are tion, to producea "staring"image as well as traditionalrange/bearing presentation.The newdigital beamformer operates on sampleddata, and createddue to the symmetryof certainloading conditions. The transition usestime delays rather than phase shifts for broadbandsignal capability. fromelevated response at a boundaryto uniforminterior response is stud- It alsofeatures dynamic, pulse-tracking focusing from within one aper- ied asa functionof theratio of bandwidthto centerfrequency •'h/f• ), for ture of the array. An experimentalimaging sonar with a 64-channeldigi- combinationsof 1-, 2-, and 3-D modesemanating from corners,edges, tal beamformerhas been constructed. A highlyparallel, pipelined archi- and walls (in the caseof a rectangularenclosure). The sizeof the transi- tectureis employed to obtaina processingrate of four million beam output tion regionseales with the centerfrequency and is affectedby the band- widthto centerfrequency ratio in varyingdegrees depending upon mode samplesper second, delayed and summed across 64 hydrophonechannels. type and geometry.In the AMA limit, the intensificationzone can be This paperdescribes the technology and presents results obtained with the describedby a localsolution that is independentof the enclosuredimen- prototypesystem, including samples of imagery. sions.Wall absorptionaffects the characterof the intensificationzone, and this effectis studiedfor variousf•/f• ratios.Local solutionsformed from the superpositionof obliquewaves provide the formulationfor inten- siftcationbehavior. By expandingthe resultin a Taylor seriesabout the 2:15 centerfrequency, the intensificationzone can be describedby a set of universalcurves that containthe center frequency effect and a bandwidth H6. A high-frequencyomnidirectional source for a robot end effector. correction.• Work supportedby NASA LangleyResearch Center. ] Jack Leifer and Ilene J. Busch-Vishniac(Department of Mechanical Engineering,The Universityof Texasat Austin,Austin, TX 78712) A high-frequencyomnidirection source has beendesigned that en- ablesthree-dimensional position monitoring in air to be performed.The 3:00 sourceconsists of a piezoelectricdisk sandwichedbetween the largeends H9. Somecalibrations and applicationsof the aerodynamicfan reference of two solidexponential Webster horns. When appropriately housed and sound source.Angelo J. Campsnells(Campsnells Associates,3201 drivento longitudinalresonance, the narrowend of oneof the hornscan RidgewoodDrive, Columbus,OH 43026) bemodeled as a baffledpiston. Because of the highdriving frequency { 301 kHz) andthe small radius of thehorn, the directivity pattern of the source Introducedby H. C. Hardy [Noise Control (May 1959)], the me- containsno nodes.Preliminary data regardingthe soundfield produced chanicalreference sound source (RSS) providesa broadbandsound field by the sourcewere presented previously [J. Leiferand I. J. Busch-Vish- by turbulentair shearat the perimeterof a rotatingopen centrifugal fan nice,J. Acoust.Soc. Am. Suppl.I 85, S115(1988) ]. Sincethat time,a wheel.For a givenwheel rpm, geometry,and air deusity,the emitted morequantitative study of thesource field has been performed, the results soundlevel is constantwithin lessthan I dB when averagedover a few of whichare presentedhere. It wasfound that the hornsdid not resonate seconds.The soundpower is relativelyuniform in octaveor third-octave exactlyat the theoreticallycalculated frequency so a mechanicaltuning bands from 125 to 8000 Hz. Usable levels of sound power are emitted processhad to be employed.The predictedand achievedresonant fre- aboveand belowthese frequencies, with field measurementsin the 63-Hz quencyof the sourcediffers by about7%. [Work supportedin part by octaveband feasiblein large rooms.Small motor-drivenunits have en- General Motors ResearchLaboratories. ] dured for decades. Hemi-anechoic, reverberation room and intensity methodsJR. Peppinet aL, NOISE-CON 88, and W. D. Gallagher(pri- vate communication)] for RSS calibration will be described. Low-fre- quencydiscrepancies between free-field and reverberationroom calibra- tion methods[ P. Brnel, B & K Review ( 1978,No. 3) ] havediminished as 2:30 thereverberation room method is improved (better sound diffusion, space averaging,and decayanalysis) [P. K. Baade, 1989 ASHRAE Trans. H7. Self- and mutual radiation resistance measurement by use of the (January 1989) ]. The portabilityand reliabilityof thisdevice suit it espe- sound intensity technique. Hideo Suzuki, Masazou Anzai, cially to worldwideuse in hostileenvironments such as factoryassembly Takahiko Oho (AcousticsLaboratory, Oho Sokki Company Limited, roomsand constructionsites. Some recent experiences in applicationof 2-4-1 Nishishinjuku, Shinjuku-ku, Tokyo, 163 Japan), and theAcculab RSS to smalland large device sound power measurement and Frank Fossaert(Institut SuperieurD'electronique de Paris, Paris, roomsound absorption will be described. France) Radiatedsound power from sourcescan be measuredby useof the soundintensity technique. The radiatedpower is obtainedby integrating the normal intensity over the surface surrounding a sound •ourcc. Since 3:15 the total radiatedpower from soundsources is determinedby self- and mutualradiation resistances, it is possible to determinethe radiation resis- H10. Characterization of the properties of underwater acoustical tancesfrom the measurementof radiatedpower. For a soundreinforce- materials.Robert Y. Ting (UnderwaterSound Reference Detachment. mentsystem, it is usefulto know the radiationresistances of loudspeakers U.S. Naval ResearchLaboratory, Orlando, FL 32856-8337) S21 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S21 Increasedemphasis on thedevelopment of newmaterials for under- Comparisonsof diffraction discrimination bya hydrophonearray versus wateracoustical applications requires that materialproperties be careful- a singlehydrophone at various positions are presented. ly characterizedduring the research phase of the investigation.Once mea- suredin the laboratory,the chemical,physical, and acousticalproperties of thematerials are correlated in orderto enableone to designnew acous- tical materials.This approachhas been successfully employed at NRL- 3:45 USRD for the pastyears in the studyof bothnew transduction materials and acousticalelastomers. Piezoelectric and dielectricproperties of new H12. A generalizedapproach to r•mdom noise synthesis:Hardware ceramicsand piezocompositesare characterized.Dynamic moduli in implementationand performsracecharacteristics. K. D. Hsueh and R. shear, tensile, and bulk modes are also determined for elastomers,while P. Hamernik (AuditoryResearchLaboratory,107 Beaumont Hall, State theirchemical properties are analyzed by usinga numberof instrumental Universityof New York, Plattsburgh,NY 12901) techniques,including gel permeationand liquid chromatography, Four- A theoreticalmodel and a computersimulation of a generalizedap- ier transforminfrared spectroscopy, and thermal gravimetricanalysis. proachto the synthesisof randomnoises were presentedin an earlier Thesemeasurement techniques will be described,and their frequency, paper[K. D. Hsuehand R. P. Hamernik,J. Acoust.Soc. Am. Suppl.1 85, temperature,and pressure capabilities will bediscussed. S40 (1989) ]. In this paper,the hardwareimplementation for this new synthesistechnique is described.The digitallygenerated analog signals transducedthrough a high-intensityacoustic driver show that manipula- tionsin thephase domain can be advantageous in creatingacoustic signals of widelyvarying descriptors (e.g., skewhess, kurtosis, etc. ) while main- tainingidentical amplitude spectra. In essence,entire families of random 3:30 waveformhaving identical amplitude spectra but variablestatistical prop- Hll. Diffraction about acoustically soft panels. Ronald P. Radlinski ertiescan be created.The influenceof the acousticdriver on the spectra (Naval Underwater SystemsCenter, New London, CT 06320) and the statisticalproperties of the synthesizedwaveforms will be shown. Someexamples of the flexibility of phasemanipulations will be presented, The useof two-dimensionalshaded and unshadedarrays to discrimi- e.g.,( 1) amplitude-modulatedrandom phase and (2) convolutionsof the nate againstdiffraction about acoustically soft panels is investigatednu- synthesizedwaveforms with any arbitrary function.These resultsillus- mericallyby the use of a two-dimensionalboundary element integral trate that a varietyof phasemanipulations can be performedto produce method.The calculatedpredictions for an infinitelylong strip are usedto interestingfamilies of complexrandom waveforms. The applicationof estimatediffraction at a hydrophonearray behind a squarebaffle. The this methodologyto acousticor vibrationtesting, environmental stress diffractionfloor for insertionloss of hydrophonearrays is determinedby screening,psychoacoustic studies, etc., will bediscussed. [Work support- weightingthe hydrophoneresponses with variousshading coefficients. ed by NIOSH.] S22 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S22 TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM B, 1:00TO 5:00 P.M. SessionI. PsychologicalAcoustics II: Memorial Sessionfor JamesP. Egan Neal F. Viemeister, Cochairman Departmentof Psychology,University of Minnesota,75 East River Road, Minneapolis, Minnesota 55455 Dennis McFadden, Cochairman Departmentof Psychology,Mezes Hall, The Universityof Texas,Austin, Texas78712 Chairman's Introduction•l:00 ContributedPapers 1:05 Recentresearch on temporalresolution indicates that a personcan listenusing a temporalwindow with a timeconstant of lessthan 10ms. A I1. Respect for the variables--•ames P. Egan remembered.James basicquestion is how informationfrom thesebrief "looks" is combined D. Miller (Central Institute for the Deaf, 818 S. Euclid, St. Louis, MO and processed.The initial focusis on theold problemof temporalintegra- 63110) tionand data are reportedfrom an experimentsimilar to Zwislocki'stwo- In Egan'slaboratory a varietyof scientificeccentricities were forgiv- click experiment [J. Zwislocki, J. Acoust. Soc. Am. 32, 1046-1059 'able or evendesirable, but therewas an absoluterequirement that one ( 1960}!. The data are well describedby assumingtrue integrationfor "respectthe variables."Respect for the variablesincluded their careful briefclick separations and by assumingtwo independentlooks for separa- measurement,accurate recording, their unconfoundedcontrol, and be- tionslonger than approximately 10 ms. Thus thedata are more consistent yond that a certainreverence. Egan's precept of "respectfor the vari- with multiplelooks than with true, long-durationtemporal integration. ables"has had a specialimpact on hearing-aidresearch at CID. Since To accountfor moregeneral "integration" phenomena, a simple multiple- 1966 with the first attemptsto understandwhy a simplemicrophone- lookstrategy is compared with neuralsummation using spike data record- amplifier-earphonecombination performed better than a hearingaid ed from the auditorynerve ofchinchilla for CF tones.The decisionstatis- [Miller andNeimoeller, J. SoundHear. Res.(1967) ] to thecurrent work tic is either an optimally weightedcombination of the spikecount from on fittingdigital hearing aids, this work hasbeen guided by the "respect- contiguousshort bins (looks) or the total spikecount summed over the the-variables"precept. Methods developed at CID andelsewhere have led durationof the tone.The time-intensitytrading functions (threshold for to increasingaccuracy in specifyingthe acoustic performances of hearing d' = I versusduration} for theseschemes are very similar. This demon- aidsin relationto the listener'sauditory area. Progress and current issues strates,consistent with intuition, that a multiple-look schemecan show will be reviewed. integration-likebehavior and that suchbehavior does not implytrue inte- gration. [Work supportedby NS12125.] 1:20 12. Prolmrtional target-tone duration as a limit on pattern discriminability:Multicomponent targets. Charles S. Watsonand Gary R. Kidd (Department of Speech and Hearing Sciences,Indiana 1:50 University,Bloomington, IN 47405) Underhigh levels of stimulusuncertainty, pattern discrimination per- 14. Recognition memory for arbitrary, complex waveforms. William formanceis often degradedso severelythat it would be reasonableto A. Yost (ParrelyHearing Institute, Loyola University, 6525 N. Sheridan questionwhether the sameprocessing mechanisms used under minimal- Road, Chicago, IL 60626) uncertaintyconditions are still available to the listeners.It haspreviously beenreported that the degradingeffects of highstimulus uncertainty de- pendstrongly on the proportional duration of a patterncomponent that is Recognitionmemory for arbitrarycomplex stimuli was measured us- subjectto change,relative to the totalduration of thepattern [1. Acoust. ingthe operating characteristic technique developed by Egan[ J. P. Egan, Soc. Am. Suppl. I 83, S40 (1987)], but only weakly on the absolute AFCRC TN 58-51,AD 152650(1958) ]. Ten complexstimuli (500 msin durationsof the stimuli.Again usingtonal sequences and a frequency duration}, eachconsisting of six,randomly chosen, equal amplitude sinu- discriminationtask, this findingwas extendedto the caseof multiple- soidalcomponents spanning frequencies from 300 to 3000Hz, wereused componenttarget sounds.Results suggest that detectabilityof pattern asa trainingset. After 15rain of listeningto thetraining set, listeners were changesis a functionof both the proportionalduration of eachchanged presenteda 20-stimulustest set consisting of theten training stimuli along tone and the numberof changedtones, and not solelya functionof the with ten additionalcomplex stimuli (generatedin the sameway asthe ten total proportionof the patternthat hasbeen changed. [Work supported training stimuli). In a recognitionmemory paradigm, listeners used a by AFOSR and NIH. ] five-pointrating scale to rate their confidencethat a stimuluspresented from the test set was from the trainingset. Different constraintswere placedon the selectionof the complexwaveforms to determinehow those constraintsaffected recognition memory. For onecondition there were no 1:35 constraintsexcept as described above; for a secondcondition the six-tonal componentswere harmonics(randomly chosen) of a commonfunda- 13. Multiple looks and temporal integration. Neal F. Viemeister mental;and for a third conditionthe six-tonecomplex was sinusoidally (Departmentof Psychology,University of Minnesota,Minneapolis, MN amplitudemodulated at ratesof 4, 16, or 32 Hz. The resultswill be dis- 55455) and Gregory H. Wakefield (Department of Electrical cussedin termsof auditoryobject perception and the useof receiveroper- Engineeringand Computer Science, University of Michigan,Ann Arbor, atingtechniques to measurerecognition memory. [ Work supportedby a M! 48109) ProgramProject Grant from the NINCDS and a grantfrom AFOSR. ] $23 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S23 2:05 2:20 I5. The detection of in-phase and out-of-phasetones presentedeither 16. Spectral interference and the masking-level difference. Leslie randomly intermixed in the sameblocks of trials or in separateblocks of R. Bernstein (Center for Neurological Sciences/SurgicalResearch trials. David S. Emroerich, William S. Brown, and James B. Thorns CenterUniversity of ConnecticutHealth Center,Farmington, CT 06032) (Department of Psychology,State Universityof New York at Stony Brook, StonyBrook, NY 11794-2500) Severalinvestigations suggest that sensitivityto changesin interaural When signalsthat are binaurally in phaseare presentedin a back- disparitieswithin selectspectral regions may be degraded by the presence groundof in-phasenoise, they are generallyconsidered to be detectedby of energyat other,even remote, spectral regions. tt seemedinteresting to meansof monauralprocessing. However, when out-of-phase signals are determinewhether similar degradations would be observedin an MLD presentedin thesame background 0fin-phase noise, binaural processing is paradigm.Detection thresholds were measured for NoSoand NoS•. The involved.One questionaddressed in the presentresearch is whethersub- signal,and 800-Hz tone (100 ms), was presentedin continuous,broad- jects are ableto processsuch monaural and binauralsignals as efficiently bandnoise. Thresholds were also measured in the presenceof an antipha- when they are randomly intermixed within the sameblocks of trials of a sic,400-Hz tone (the interrefer)gated simultaneously with the signalor single-intervalyes-no procedure, as when they are presentedin separate presentedcontinuously. NoS•- thresholdsincreased by about 7 dB with blocksof trials. As in other experimentsinvolving stimulus uncertainty, the gatedinterferer at 80 dB SPL. Smallerincreases were observedwith evidencewas found of some impairment associatedwith listeningfor lower levelsof the interferer. Slightly increasingor decreasingthe fre- monauraland binaural signalssimultaneously. Response latencies were quencyof the interrefer awayfrom the octavereduced substantially its alsoexamined in order to gainadditional information about stimulus pro- effect,as did presentingit continuously.NoSo thresholdswere affected cessing.Latency-based operating characteristics were derivedusing the onlyslightly by the interferer.Reversing the rolesof thesignal and inter- sameprocedure employed by Emmerichet aL [ Percept.Psychophys. 11, ferer (400-Hz signal,800-Hz interferer)led to smaller,but reliabledegra- 65-72 (1972) ], amongothers, and comparedfor monauraland binaural dationsin performance.Homophasic interferers had, in general,smaller listeningconditions. effectson performance.[Work supportedby NIH andAFOSR.] 2:35-2:45 Break 2:45 the psychometricfunction is not critical in achievinghigh efficiencies. Finally, data from human listenersindicate that the standarddeviation of I7. How onset,offset, and ongoinginteraural delaysaffect lateralization. thresholdestimates is approximatelytwice as large when the stimulus ThomasN. Buell and ConstantineTraMotis (Center for Neurological presentedon each trial is at a stimuluslevel correspondingto 70.7%, Sciences,Surgical Research Center and Department of Surgery rather than 94% correct. [This researchwas supportedby NSF and (Otolaryngology), University of Connecticut Health Center, AFOSR. ] Farmington,CT 06032) Severaltypes of interauraldelay can affectthe lateralposition of bin- 3:15 aural signals.Delays can occur within the transient (onsetand/offoffset) or steady-state(ongoing cycle-by-cycle and/or envelope-based) portions 19. Psyehoaeoustiealmasking of one pure toneby anotherand its relation of the signal,or both. The extentof lateralityproduced by thesedelays for to a model of the mechanics of the basilar membrane. Julius L. Goldstein simple and sinusoidallyamplitude-modulated tones with an acoustic (Central Institute for the Deaf, 818 S. Euclid Avenue, St. Louis, MO pointingtask were measured. Relative potency of thesedelays was as- 63130) sessedby presentingthem singlyor in combinations(where the delays wereconsistent or in opposition).Also varied were rise/decay time, dura- Acceleratedpsychoacoustical masking effects with levelfor simple tion, rateof modulation,and centerfrequency. The majorfinding was that tonemaskers of higherfrequency tones [Wegel and Lane,Phys. Rev. 23, ongoingdelays are muchmore potent than shownin previoustasks that 226-285 (1924); Duifhuis, J. Acoust. Soc. Am, 67, 914-927 (1980)] requireddiscriminations rather than judgmentsof location. [Research haveclear correlates with nonlinearsuppression and excitation phenome- supportedby NIH and AFOSR. ] na in auditorynerve and basilarmembrane responses [Abbas and Sachs, J. Acoust. Soc.Am. 59, 112-122 ( 1976); Patuzzi, Sellick,and Johnstone, Hear. Res. 13, 19-27 ( 1984); Robles etal., $. Acoust. Soc. Am. 80, 1364- 1384 (1986) ]. A signalprocessing model ofbasilar membrane response is presentedthat quantifiesbasilar membrane mechanics as two nonlinearly 3:00 interactingfilter responses:(i) a sensitive,narrow-band compresslye re- I8. Optimal stimulusplacement in adaptivepsychophysical procedures. sponseresponsible for thetips of cochlearsuppression and excitation tun- David M. Green (PsychoacousticLaboratory, Department of ingcurves; and (ii) andinsensitive, lowpass linearlike response responsi- Psychology,University of Florida, Gainesville,FL 32611) ble for the low-frequencytails of tuning curves.Dual filter processing alongwith mutual suppressionbetween the filters,described by the mod- A commonmisconception is that the stimulusshould be presented el, allowsconceptualization of knownpsychoacoustics in termsof com- nearthe threshold value in an adaptivepsychophysical procedure to mini- petingcochlear mechanical responses. Detailed model accounts of data mizethe variance of thresholdestimates. For maximumlikelihood proce- andmethods for estimatingmodel parameters will be presented. dures,recent work by Laming,Watson, and Pelli showsthat the stimulus should be presentedat a relatively high level. For some psychometric functions,they found the optimalstimulus placement level produces 84% 3:30 to 94% correctin a two-alternativeforced-choice task. This resultis par- ticularly disquieting,because the populartwo-down one-up rule tracksa I10. A psyehophysicalevaluation of a model for suppressionand relativelylow percentageof correctresponses, 70.7%. Our computersim- excitation in remote masking. J. L. Goldstein, R. H. Gilkey, and R. ulations,using a varietyofpsychometric functions, confirm the validityof E. Quifi6nez (Signal Detection Laboratory,Central Institute for the theiranalysis. These simulations also demonstrate that theprecise form of Deaf, St. Louis, MO 63110) $24 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S24 In theclassical literature both the masking and the suppression of one underpower-law ROC curvesis simply1/( I + k). An alternativeap- toneby a secondtone of lowerfrequency have been shown to benonlinear proachis to assumean ROC curvewith linearsegments between 0,0 and functions of overall level. Here, the relation between these two tasks is H, FA; andbetween H, FA and 1.0,1.0. The calculated area by thelinear furtherconstrained by measuring both remote masking and suppression assumptionis notconstant for a fixedROC curve;it varieswith the H,FA with the samesubjects. The remotemasking experiment measures the level.The linear assumption underestimates, and at timesseriously under- effectofa 602-Hz sinusoidalmusket on the detectabilityof a simultaneous estimates,the area under power-law ROC curves.The minimumerror of 1500-Hzsinusoidal signal, while the suppression experiment measures the underestimation is over 4% for k between 0.05 and 0.5. The maximum effectofa 602-Hzsinusoidal suppressor on theforward masking produced errorof underestimationapproaches the maximum possible error. Clear- by a 1500-Hzsinusold. The dataagree with earlierstudies and are inter- ly, thelinear approximation procedure can grossly distort the area under pretedin the contextof the multipleband pass nonlinearity model [J. L. the ROC curve. Goldstein,"Updating cochlear driven models of auditoryperception: A model for nonlinearauditory frequency analyzing filters," in Working Modelsof Human Perception,edited by B. Elsendoornand H. Bouma (Academic,New York, 1989),pp. 19-57], whichsuggests that bothexci- II2. Spectral tuning for a temporal decline of masking. tatory and suppresslyemechanisms influence remote masking. [Work Dennis McFadden(Department of Psychology,University of Texas, supportedby NSF and AFOSR. ] Austin, TX 78712) When a tonal signalof 1250-Hz and SO-msduration is maskedby a spectrallynonoverlapping, narrow-band noise centered at 2050 Hz, the amountof maskingobserved depends upon the durationof thisflanking 3:45 noiseband. Specifically, when the flankingband is gatedon onlyfor the Ill. Area under power-law receiver operating characteristics. durationof theobservation interval (burst), detectabilitycan be 20-25 dB Irwin Pollack (Mental Health Research Institute, University of worsefor somesubjects than whenthe flankingband is gatedon a few Michigan, Ann Arbor, MI 48109-0720) hundredmilliseconds prior to the observationinterval (fringed). This temporaldecline of maskingas the fringeduration increases has the ap- The receiveroperating characteristic, or ROC curve,describes perfor- pearanceof an adaptationeffect---or what Egan used to call perstimula- manceunder a widerange of hit (H) andfalse alarm (FA) probabilities. tory fatigue.When the fringeconsists of a singletone of 2050Hz preced- In actual practice,with many differentexperimental conditions, only a ing the flankingband, a substantialtemporal decline of maskingcan still singlehit-false alarm (H-FA) combinationmay be availableto describe be obtained.As the frequencyof this tonal fringe is variedaround 2050 performanceunder each condition.One way to describeeach perfor- Hz, the magnitudeof the improvementrelative to the burst condition mancelevel is in termsof thearea under an ROC curvepassed through the diminishes,and, at leastfor mostlisteners, a "tuned"existence region can obtainedH-FA combination.lames Egan astutelyobserved that most be mappedfor this temporaldeline of masking.Possible explanations ROC curvesare well approximatedby power-lawfunctions of the form: includetemporal development of thecritical-band filter, and within-chan- H = [ FA] s.Egan's power-law functions also satisfy the requiremere of nel adaptationfollowed by across-channelmasking. [Work supportedby passingthrough the coordinatepairs: 0,0; H-FA; and 1.0,1.0.The area NINCDS. ] 4:15-5.-00 OpenDiscussion TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM A, 1:00 TO 2:40 P.M. SessionJ. SpeechCommunication II: Focus Sessionon Creaky Voice Caroline G. Henton, Chairman LinguisticsProgram. University of California,Davis, California95616 Invited Papers 1:00 Jl. Physiologyof creaky voiceand vocal fry. Ronald C. Scherer(Recording and ResearchCenter, The DenverCenter for the PerformingArts, Denver,CO 80204) Creaky voiceand vocalfry are voicequalities that are linguisticallynormal but oftensymptomatic in voice pathology.The physiologyof creakand vocalfry will bediscussed from the standpointsof electromyography, videofiberoscopy,x-ray, glottography,and aerodynamics.Discussion will emphasizethe relativerole of the cricothyroidand thyroartenoidmuscles, laryngeal vestibule constriction, hyperadduction, vocal fold motion irregularity,subglottal pressures, mean flow, and glottalsource flow. New data in theseareas will becompared with earlierfindings. [Work supportedby the BonfilsFoundation. l S25 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America $25 1:25 J2. Acousticsof creakyvoice. Ingo R. Titzc (VoiceAcoustics and BiomcchanicsLaboratory, Department of SpeechPathology and Audiology,The Universityof Iowa, Iowa City, IA 52242 and the Recordingand ResearchCenter, The Denver Center for the PerformingArts, Denver, CO 80204) Creaky voiceseems to be perceivedas somecombination of low pitch and roughness.Preliminary percep- tual experimentssuggest that wheneverthe fundamentalfrequency drops below about 70 Hz, pulsesare perceivedin phonation.The combinationof severalclosely spaced harmonics of different (and varying) amplitudesmay then add roughnessifthe harmoniesfall within the samecritical band. In additionto relaxing the laryngealmuscles to producelow Fo, the larynx can ereateasymmetric vibrations and modechanges that producesubharmonic components. These contribute to the perceptionof low F<•.Subharmonics are a resultof bifurcationof a singlestable vibratory statc into multiplequasistable states and can Icadto chaoticvibratory behaviorif asymmetriesor nonlinearitiesin the systemare increased. 1:50 J3. Sociophoneticaspects of creaky voice. Caroline G. Henton (LinguisticsProgram, University of California, Davis, CA 95616) In a wide-reachingstudy of creakyvoice in BritishEnglish (Henton and Bladon,1988) it wasindicated that useof creakyvoice varied significantly according to sexand accent of speakers,as well aswith utterance position.Specifically, males used between three and ten timesas much creaky voice as did females.Accentual differenceswere alsofound: Speakers of ModifiedNorthern were more polarizedin the overallamounts of creakused than were RP speakers.Individual speakers varied considerably in theiruse of creak,with a number of malesbeing persistent creakers. In addition,creaky voice was shown to bean "end-of-utterance"phenome- non,used by femalesand malesin a demarcativefunction. Further researchusing data from AmericanEnglish and from Frenchindicates that creakyvoice may be employedas a sociophoneticmarker across cultures. Indeed,its roleas a markerof malespeech may be enhanced in certaincultures where gender-role expectations are morepronounced. The demarcativeintonational function is observedagain cross linguistically. This phe- nomenon,coupled with Hyman's (1988) recentresults for West African languages,may make intonational creakyvoice a candidatefor a phoneticuniversal. 2:15 J4. Cross-linguisticrelations between voice quality and vowel quality. Keith Denning (Department of EnglishLanguage and Literature,Eastern Michigan University, Ypsilanti, M I 48197) A surveyof impressionistieand instrumentaldescriptions of diverselanguages (including new data from the Nilotic languagesDinka and Acholi) revealsrepeated cases of phonetalcand subphonemicinteraction betweenpronation type and aspectsof oropharyngealresonance. Generally, if thereis a regularinteraction betweenintra- and extralaryngealfactors in a language,vocal {'oldconfigurations with greaterlaryngeal laxness(as requiredfor regularphonation in consonants,murmur in consonantsand vowels,and pitch depres- sion) are associatedwith lower F 1 and, conversely,greater vocal fold tension(associated with voicelessness, creakyphonation, and higherpitch) occurswith higherF 1. Explanationsfor thisuniversal in biomechanical (Laver, 1980), neuromuscular( Laufer and Condax, 1981 ), andaerodynamic ( Oh ala, 1983) aspectsof articu- lation are proposed.Cross-linguistic evidence suggests that articulationfrom intra- and extralaryngealdo- mains(putatively separate for phonologiealpurposes) may functiondistinctively only in combinationwith eachother in manycases. Implications of theseand otherfindings include the needfor a distinctivefeature of "complex voice quality," in contradistinctionto the "primary feature" approachof Stevens,Keyser, and Kawasaki(1986), and the desirabilityof phonologicalmodels that accommodatephonetically based patterns of the relevanttype in formulatingprinciples of soundchange and synchronicphonological analysis. S26 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S26 TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM G, 1:15 TO 5:15 P.M. SessionK. PhysicalAcoustics II andBioresponse to VibrationI: BiophysicalMechanisms as Bases of Safety Criteria for Medical Ultrasound I WesleyL. Nyborg, Chairman Departmentof Physics,University of Vermont.Burlington, Vermont05405 Invited Papers 1:15 KI. Chairman's introduction:Mechanisms as bases for safety criteria. Wesley L. Nyborg (Physics Department,University of Vermont, Burlington,VT 05405) There is now considerableinterest around the world in developingscientifically based safety limits for medicalultrasound, especially as applied in diagnosis.While there has been no evidence of harmcaused by past exposuresto diagnosticultrasound, some of the newerapplications employ higher intensities than wereused formerly.Hence, there is need for quantitative information on acoustic conditions that pose different degrees of risk,to guidemanufacturers, users, and regulators. It is fairlygenerally agreed that a promisingway to generate suchinformation is to considerthe separate biophysical mechanisms by whichultrasound can cause biological damage.This specialsession is to deal with researchfindings that relateto thesemechanisms, and their relevanceto safetylimits. 1:30 K2. Nonlinearaspects of ultrasonicheating. E. L. Carstensen,D. Dalecki,K. J. Parker(Department of ElectricalEngineering, Biophysies, and Radiology, and the RochesterCenter for BiomedicalUltrasound, Universityof Rochester,Rochester, NY 14627),R. Bacon(Division of RadiationScience and Acoustics, NationalPhysical Laboratory, Teddington, Middlesex TWI 1 0LW, UnitedKingdom), and D. Blackstock (Departmentof ElectricalEngineering, Unversity of Rochester,Rochester, NY 14627and Department of MechanicalEngineering and Applied Research Laboratory, University of Texas at Austin, Austin, TX 78712) Underappropriate conditions, the absorptionof finiteamplitude ultrasound is determinedlargely by sourceamplitude, field geometry, and the nonlinearproperties of themedium and is onlyweakly dependent uponthe small signal absorption coefficient ofthe material. The characteristics ofthis absorption are complete- ly foreignto theideas that we have come to associatewith small signal losses. Under certain conditions, finite amplitudeabsorption is somewhatgreater in focusedthan in plane-wavefields. Under medically relevant conditions,heating of tissuesmay be enhancedby a factorof 3 or moreby nonlinearphenomena. Some commerciallyavailable diagnostic instruments may be capableof producingfocal pressure amplitudes high enoughto resultin nonlinearcontributions to localheating. The resulting total temperature increments could exceedthe I 'C, the valueuniversally regarded as safe.Finite amplitude processes affect the thresholdsfor tissuedamage. The characteristics ofthe threshold curves differ qualitatively depending upon whether they are expressedin termsof localor sourceintensities. [This work was supported in partby U.S.P.H.S.Grant No. CA39241. ] K3. Biophysicalcharacterization of diagnosticultrasound equipment. Gall ter Haar (PhysicsDepartment, F Block,Institute of CancerResearch, Sutton, Surrey, United Kingdom) and FrancisDuck (MedicalPhysics Department,Royal United Hospital, Bath, Avon, United Kingdom) Concer•for thesafety ofdiagnustic ultrasound procedures has meant that therehas been a moveto classify ultrasonicscanning equipment in termsof the physicaleffects it can producein biologicaltissues. To this end, differenttypes of diagnosticequipment currently in usein the clinic havebeen characterized in termsof their heatingeffects, thor abilityto indueecavitation bubbles in agargels, the streaming velocity attained under free- field conditionsin water, and their ability to producesonoluminescence. The occurrenceand extent of these phenomenahave then been studied in the lightof theacoustic parameters (such as peak positive pressure, peak negativepressure, spatial peak temporal average intensity, etc.) usedto producethem. Extensive characteriza- tionof themachines investigated was carried out usinga beamca]ibrator. It wasfound that noneof theacoustic parametersmeasured provided an unambiguousindication of the extentto whichany of thesephysical mecha- nismswould occur, thus rendering it difficultto predictsafety on thisbasis from a listingof acousticoutput. S27 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S27 2:20 K4. In vitroultrasonic heating of fetalbone. J. L. Drewniak,K. I. Carnes,and F. Dunn (Bioaeousties ResearchLaboratory, University of Illinois,1406 West Green Street, Urbana, 1L 61801) Estimatesof temperature increase in soft fetal tissues exposed to medicalultrasound can be made with existingknowledge. However, little knowledge isavailable tomake a resonableestimate ofthe temperature increaseinfetal bone and surrounding tissues during in uivoexposures. An experimentalstudy of the tempera- tureincrease measured in vitro in humanfetal femurs exposed to I-MHz, continuouswave ultrasound at 37 øC hasbeen conducted. The temperature increase in thefemur specimens ismeasured with a smallimplanted thermocoupleprobe. Specimens range in gestational age from 59 to 150days. The temperature increase mea- suredina 59-day gestational agespecimen at20 s following theinitiation ofexposure at1 W/cm 2, for example, was0.10 øC, while the temperature increase for a 108-dayspecimen with the same exposure conditions was 2.9øC. A temperatureincrease of4.0 øC was measured inthe 108-day specimen fora 60 s exposure at1 W/cm •. Thetemperature riseis found to increase linearly with intensity over the range of 0.5 to 10 W/cm 2. The initial rateof the temperature increase, evaluated at0.2 s following initiation ofexposure, has also been measured and isfound to increaserapidly with gestational age. For example, I • dT/dt at 0.2s measuredfor the 59-day specimenis0.06 øC/s/(W/cm2), while for the 108-day specimen 0.81 øC/s/(W/cm 2) was measured. The valuesof initial temperature increase measured inthe fetal femur specimens can be compared to thatobtained forsoft tissue using the quantity I - • dT/dt= 2a/pCp.For a = 0.05cm- •, andpCp = 3.78$/•C/cm •, I - • dT/dt = 0.026øC/s/(W/cm2). The values ofI - •dT/dt measuredfor the 59- and 108-day gestational age specimensare approximately 2 and 30 timesgreater than that for softtissue. 2:45-3:00 Break 3:00 KS. The potentialrole of acousticcavitation in roedialultrasound bioeffects. L. A. Crum, R. A. Roy, and C.C. Church (National Centerof PhysicalAcoustics, University of Mississippi,University, MS 38677) Ultrasoundis usedin a varietyof modalitiesin medicine.Some of theseuses are: therapeutic--in which ultrasoundis intended to havea (beneficial)biological effect; mechanical--in which vibrating tools are used as surgicalscalpels, as dental scalers, etc.; diagnostic--in which clinical evaluations are made of internalorgans andbody conditions; clinical--in whichhigh-intensity shock waves are used to cornminuteundesirable concre- tionssuch as renal or biliary calculi.For all thesemodalities, when their waveformsand typical acoustic intensitiesare reproducedin aqueousliquids in the laboratory,acoustic cavitation normally occurs. Since cavitationscan also result in severemechanical damage to objectsin its vicinity,its presencein vivois therefore to be avoided.This grouphas been active for someyears in the studyof cavitationproduced by medical ultrasounddevices and reports will bepresented on variousaspects of theseinvestigations. [ Work supportedin partby NIH GrantNo. 2 R01CA 39374-04AI, andby theOffice of NavalResearch.] 3:25 K6. Acoustic scattering from transient, micron-sized cavitation bubbles. Robert E. Apfel, SameerMadenshetty, Ronald A. Roy, and Qihong Xu (Departmentof MechanicalEngineering, Yale University,New Haven,CT 06520) Ifcavitationoccurs in vivodue to theapplication of medicalultrasound, it islikely to bedifficult to observe. Concernrests less with the destructionof biologicalcells and more with the moresubtle subcellular injuries that couldresult in damageto componentsresponsible for cellreproduction. Cavitation on the scaleof single biologicalcells--microcavitation---can be observed in vitroby scatteringhigh-frequency ultrasound (30 MHz) off of singletransient cavitation bubbles. An activecavitation detection (ACD) hasbeen used to determinethe onsetconditions (e.g., cavitation thresholds) in vitrofor waterwith variousartificial eavitation nuclei.More recently,the samescattering technique has beenemployed to observethe radiusversus time characteristicsof transientcavitation bubbles of oneor morecycles. Reports on boththreshold measurements andattempts to measurequantitative radius versus time histories of individualbubbles will bepresented. The resultswill be presentedin the generalcontext of thejoint Yale-NationalCenter for PhysicalAcoustics collaborativeeffort to assessthe safety of diagnosticultrasound. [Work supportedby theNational Institutes of Health throughGrant 4R01-CA39374.] 3:50 K7. Implicationsof gas body activationfor medicalultrasorties. Douglas L. Miller (Battelle Pacific Northwest Laboratories, P.O. Box 999, Richland, WA 99352) A stabilizedbody of gasin a biologicalmedium can be directlyactivated (set into oscillation)by mega- hertz-frequencyultrasound, and cause nonthermal bioeffects at levelsbelow the cavitation threshold. This phenomenonhas been studied in planttissue with intercellular gas-filled channels, in insecttissue with gas- filledtracheal systems, and in cell suspensionsprovided with gastrapped in hydrophobi½micropores. The highlylocalized oscillation leads to the gathering of particlesby radiation forces and to damaging shear stresses fromacoustic microstreaming flow. In tissues,the bioeffects seem to havediscernible thresholds and to depend onthe temporal-peak intensity of exposure.In cellsuspensions, for whicheffects involve collection of cells,a S28 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S28 dependenceon the squareof the temporal-averageintensity is discernible.In thesemodel systems, significant effectscan be observed with exposureparameters similar to thoseencountered in diagnosticmedical applica- tions.Furthermore, studies of decompressionsickness and cavitationindicate that stablegas bodies may sometimesexist in mammaliantissue. Although further study of possiblehazards is clearlyneeded, the exis- tenceof eventhis potential risk may justify concern about frivolous exposures with medicaldevices. [ Work supportedby PHS grantCA42947 awarded by the NationalInstitutes of Health.] ContributedPapers 4:15 The dramatic increasein the amountof cell lysisobserved when the exposurechamber is rotatedhas been shown to be the resultof recycling KS.U!trasonieheatingoftheskulI.EdwinL. Carstensen,SallyZ. Child cavitationbubbles through the suspension [ D. L. Miller andA. R. Wil- (Departmentof ElectricalEngineering and The RochesterCenter for liams,Ultrasound Med. Biol. (in press)]. In addition,the destructive BiomedicalUltrasound, The Universityof Rochester,Rochester, NY effects of these bubbles have been shown to be reduced when the number 14627), StephanieNorton (The RochesterCenter for Biomedical of cellswithin the suspensionwas increased, with fixedcells being much Ultrasound,The Universityof Rochester,Rochester, NY 14627), and moreinhibitory than fresh cells [ Ellwart et aL, UltrasoundMed. Biol. 14, Wesley Nyborg{The RochesterCenter for BiomedicalUltrasound, The 43-50 (1988) ]. It is proposedthat underthese exposure conditions cell Universityof Rochester,Rochester, NY 14627and PhysicsDepartment, lysisis causedby collisionbetween cells and rapidlymoving gas bubbles. Universityof Vermont, Burlington,VT 05405) If thebubble velocity is highenough then the cell is penetratedand lysed, Comparativelysimple analysis shows that diagnosticultrasound de- but this interaction also slows down the bubble. Below a certain critical vices,in somecases, may approach output levels that can produce signifi- velocityit ispresumed that the bubble attaches to (or istrapped within) a cantheating of tissues,particularly if the propagationpath includesbone. cell and ceasesto interactwith other cells.Additional support for this The modelassumes that all of the ultrasonicenergy transmitted into the proposedmechanism has been obtained from studiesutilizing erythro- bone is convertedinto heat at the surface.The heat source,therefore, is a cytesembedded in agargels and from histological observations in oivo [C. diskwith diameterequal to the soundbeam. Experimental tests of these J. Martin et al., Ultrasound Med. Biol. 7, 235-265 ( 1981) I. predictionsusing rodents show that temperatureincrements of the order of 3 øC/W/era2 areproduced in skullbone with sharply focused fields undermedically relevant conditions. [This work wassupported in part by U.S.P.H.S. Grant No. CA 39241.] 4;30 5:00 K9. Temperature rise in a focused Gaussian beam of ultrasound. KII. The study of cavitationin tissuelikegels. A. J. Mortimer, B. Junru Wu and GonghuanDu (Departmentof Physics,University of J. Trollope,and E. Birnbaum(National Research Council of Canada, Vermont,Burlington, VT 05405) Ottawa, Ontario K 1A 0R6, Canada) The steady-statetemperature elevation generated by a focusedGaus- A studyof the effectsofcavitation in agargels was undertaken as an sianbeam including the effectof perfusionhas been calculated along the extensionof studiesin this laboratoryof cavitation-inducedfree radicals beamaxis. The mediumis assumedto be a homogeneousabsorbing one. in aqueousmedia using the free radical scavenger terphthalic acid. Agar- The resultsindicate: ( 1) the temperaturerise is an increasingfunction of osegels were made at a concentrationof 1.5%agarose and 1% terphthalic the intensitygain of the focusingtransducer, but seemsnever to exceed acid in distilled,deionized waler. Gels were exposedto ultrasoundfor twiceof that at the interfaceof the transducerand the mediumgenerated timesof up to 3.5 rainin bothpulsed ( 1:1,100-1000 ms ) andcon tin uous by its unfocusedcounterpart; (2) the temperaturerise at the interfaceof modes,at intensitiesfrom I-2 W/cm-•. The liquidwas then separated the transducerand the mediumis not affectedby focusingsignificantly. fromthe gel by dialysis and the fluorescence counted using a spectrofiuor- [ Work supportedby theAmerican Cancer Society (I N- 156) andby NSF imeter.In all casesno fluorescencewas measured; however, at the higher andVermont Epscor. ] intensities,visible bubbles could be observedafter treatment. At the same intensitiesand exposure conditions, significant levels of freeradical pro- ductionwere observed in aqueoussolution and in gelsmade from pre- viouslyirradiated terphthalic acid. These results indicate that either cavi- tationthresholds in aqueousgels are significantlyhigher than would be KI0. Bubble penetrationmA major mechanismcausing cell lysis? A. expectedfrom studies in aqueoussolutions, or thatthe medium acts as an R. Williams (Department of Medical Biophysics,University of effectivefree radical scavenger, thereby limiting the possiblebiological ManchesterMedical School,Manchester M I 3 9PT, England) effect of acoustic eavitation. S29 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S29 TUESDAY AFTERNOON, 28 NOVEMBER 1989 DIRECTORS' ROW 43, 1:30 TO 4:50 P.M. Session L. Noise I: Aircraft Noise Gilles A. Daigle, Chairman Divisionof Physics,National ResearchCouncil, Montreal Road, Ottawa, OntarioKid OR6, Canada Chairman's Introduction1:30 Invited Papers 1:35 L1. Sonicboom measurements from very highaltitude and Mach numberson spaceshuttle-orbiter reentry. J. F. Stanley,E.G. Stansbery(NASA-Johnson Space Center, Houston, TX 77058), H. R. Henderson,and D. J. Maglieri (EagleEngineering, Hampton, VA 23666) STS-orbiterreentry sonic boom measurements are presented with particularemphasis on thoseacquired at very high altitude and Mach numbers.Included, also, are the orbiter physicalcharacteristics and teentry parametersalong with a discussionof the natureof its sonic-boomcharacteristics as comparedto previous experiencewith aircraft.The applicabilityof existingsonic-boom theory, used for aircraft,to the STS-orbiteris shown.Previous STS teentry boom measurements, along with recentlyacquired measurements from the STS- 26 andSTS-27 reentry are presented and discussed. The resultsindicate that there are no significant differences in sonic-boomsignature characteristics from STS-orbiterand otheraircraft. N-wave type signatures are ob- servedon STg-orbiterup to Math 23 andaltitude of 250 000 ft havingan overpressureof about0. i psfand a periodof about3.0 s. It isalso shown that orbiterteentry maneuvering conditions do not producefocus booms. AlthoughSTS-orbiter signatures are predictable up to Mach6 andan altitude of 130000 ft, sonic-boomtheory improvementsare requiredfor Mach numbersgreater than 6.0. [Work supportedby NASA andUSAF.] 2:05 L2. Weakersonic booms may be considerably more quiet. Allan D. Pierce(Graduate Program in Acoustics, PennsylvaniaState University, 157 Hammond Building, State College, PA 16804) Presentindications are that realisticdesigns of civiliansupersonic aircraft can be achieved so that thesonic boomat theground has substantially smaller pressure jumps (shocks) than was the case for previousdesigns andfor theConcorde. Unknown as yet is the extent to whichthis achievement may increase public acceptabi- lity to theintrusion of occasionalsonic boom events. The presentpaper reviews physical acoustics research that maypertain to thisquestion, and examines the magnitudeand possible definitions of risetimes of sonicboom shocks.The detailedpressure profile at a shockdetermines the high-frequency content of theboom and affects the perceivedloudness and the startle effect, with the risetime being a roughdescriptor of thishigh-frequency content.Earlier theoretical considerations had suggested that the risetime, on theaverage, should be inversely proportionalto theshock jump. Molecular relaxation effects, however, are now known to playa majorrole and leadto the intriguingprediction that risetimes for shockjumps below approximately 50 Pa are substantially longerthan one would extrapolatefrom measuredrise timesof strongershocks using the simpleinverse proportionalityrelation. The extentand nature of atmosphericturbulence effects are still incompletelyunder- stoodand stimulatefurther research.[Work supportedby NASA-LRC and by the William E. Leonhard endowmentto PennsylvaniaState University. 1 2.35 L3. Noise from low-altitudemilitary flying. Kenneth J. Pintkin (Wyle Laboratories,2001 JeffersonDavis Highway, Suite 701, Arlington, VA 22202) and John A. Molino (Tech U-Fit Corporation, 400 Madison Street, #210, Alexandria, VA 22314} Military air tacticsrely on increasinguse of high-speedlow-altitude penetration techniques. Training for thistype of operationcan take place in speciallydesignated low-altitude military operation areas (MOAs) or alongmilitary training routes (MTRs). Environmentalassessment of noisefrom these operations has become an importantfactor in the designof thesespecial use airspaces.The noiseenvironment is differentfrom commonlystudied situations such as airport or highwaynoise in urban/suburbanareas, and it is not obvious that the usualLo, analysisis appropriate.Individual events are loud (maximumlevels in the 100-115dBA range)and short ( 10dB downdurations of a fewseconds). Noise events are infrequent(typically less than 3 or 4 per day on busyroutes) and occurat randomtimes. These operations also take placein quiet rural areas.A discussionis presentedof the nature of theseoperations and the resultantnoise environment. This noiseis S30 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S30 contrastedwith well-understood noiseenvironments. Aninterim noise metric, Lanmr (monthly average, onset ratecorrected Lan ), developedfor the U.S. Air Forceis presented. A comprehensive research plan to test L d..... or replaceit witha moreappropriate metric, is discussed. 3:05--3:20 Break ContributedPapers 3:20 The effectof atmosphere-inducedrounding on theloudness of sonic IA. Voyagerflight crew hearing threshold levels resulting from 5- and9- boomwaveforms is investigated.Waveforms with varyingdegrees of roundnessrepresentative of measuredflight test data are considered. Sim- day continuousin-flight noise exposure.Mark R. Stephenson ple equationsare developedby whichthe perceivedloudness and C- (BiologicalAcoustics Branch, Armstrong Aerospace Medical Research weightedsound exposure levels of thesesonic booms can be calculated Laboratory,Wright-Patterson AFB, OH 45433-6573), Bradley directlyfrom theirwaveform parameters. This methodshows excellent L. Billings(Audiology Center of Redlands,245 TerracineBoulevard, agreementwith morecomplex loudness calculations (e.g., by Johnson Suite101, Redlands, CA 92373),and George A. Jutila(Fortuna Family MedicalGroup, 874 Main Street,Fortuna, CA 95540) andRobinson) as well as the full C-weighted sound exposure level calcu- lation.A newset of flight test data is used to correlate shock strength with Theflight crew of theVoyager aircraft was continuously exposed to a risetime. The results are compared toa semi-empiricalmodel proposed by broadbandnoise for nearly 5 daysduring a trialflight, and for over 9 days Dr. Kenneth Plotkin from earlier work. duringtheir nonstop flight around the world. Evaluation of thethreshold shiftsresulting from these exposures represents a unique opportunity to studythe effectof humanexposure to intensecontinuous noise for such 4:05 durations.Post-flight audiometry demonstrated that the 9-day flight did notresult in largerhearing threshold shifts than those following the 5-day L7. Fluctuatingpressure and thermal environments of runwayjoint seals flight.Neither crewmember incurred a permanentthreshold shift from exposedto jet engineexhaust. Timothy W. Lancey(Department of theseexposures. Mechanical Engineering, California State University--Fullerton, Fullerton,CA 92634}and Eugene E. Cooper(Naval Civil Engineering Laboratory,Port Hueneme,CA 93041) 3:35 Jet engineexhaust flow impingingon runwayjoint sealscreates an LS. Low-frequencynoise for low flying aircraft. Louis C. Sutherland aerothermoacousticenvironment consisting of fluctuatingpressure (Wyle Laboratories,128 MarylandStreet, E1 Segundo,CA 90245), (acousticaland hydrodynamic) and incident gas temperature rapidly in- KennethPlotkin (Wyle Laboratories,2001 JeffersonDavis Hwy., creasingwith time.Fluctuating pressure at thesurface of a joint sealon a Arlington,VA 22202), and Jerry Speakman(Department of the Air runwaywas measured with microphonesand low-pressuretransducers Force,AAMRL (AFSC), Wright-PattersonAFB, OH 45433) positionedalong the runwaycenterline. Temperatures were measured withthermocouples imbedded within the seal and on thesurface, along A considerationin the evaluationof aircraftnoise impact on struc- the runwaycenterline and at severalsideline distances. Pressure and tem- turesnear airportsand underlow-level military training routes is the peraturedata wererecorded during aircraft takeoff and landingopera- significanceof thelow-frequency content below 50 Hz. Basedon analysis tions.Measurements indicated incident noise pressure levels of approxi- of measurementsobtained under military training operations, a prelimi- mately 164 dB re: 20/zPa, surfacetemperatures over 155øC, and seal naryprediction model has been developed for applicationto environmen- internaltemperatures of 100øC. Fluctuating velocity of theseal was cal- tal assessmentof possibleeffects on structure.The modelis basedon the culated,treating the seal as a fluidflowing between parallel walls. Energy hypothesisthat the majorsource to thislow-frequency noise is radiation absorptionrate was found to equal850 W/m 2,4.7% of thepower avail- fromthe aircraft wake and vortices shed by the airflow over the wings or ablefrom the incident fluctuating pressure field. Thermal energy through fuselage.The resultingempirical model predicts that theaverage maxi- convectiveheat transferwas absorbed at 34 000 W/m •, 97.5% of total mumone-third octave band noise level, L, in decibels,in thefrequency energyabsorption. rangeof 5 to 40 Hz is givenby: L =481og[V/100] + 10log[A/ R 2] + 9I, whereVis the aircraft speed in kn, `4 is the wing area, and R is the slantrange. Fortuitously, the sameexpression also provides reasona- bleestimates ofthis low-frequency noise level in thesame frequency range 4:20 forhelicopters where Vis now the helical tip speedand .4 is the rotor disk L8. Environmentalimpulse sound evaluation by L DI index:Physical and area.The resulting empirical model is similar to that developed for higher subjectivetests. G. Brambilla, G. B. Cannelli (lsfituto di Acustica, frequencyairframe noise (J. C. Hardin, NASA SP-416, 1976). Further CNR, Via Cassia1216, 00189 Roma, Italy), and S. Santoboni( lstitutodi studyis required to assess the validity of this model in comparisonto more Tecnologia,UniversitA Reggio Calabria, Via Diana 4, 89100 Reggio sophisticatedmodels of low-frequencyairframe noise. Comparisons be- Calabria,Italy } tweenthe prediction model and measured data will beillustrated. [Work supportedin partby U.S. Air ForceHarry G. ArmstrongAerospace Directmeasurements of the L [•] indexwere performed by a newdigi- MedicalResearch Laboratory.] tal apparatus[G. B. Cannelliand S. Santoboni,Italian Patent No. 48694 A/87 andInternational Patent in progress]on a groupof soundshaving different impulsivity and recorded in the environment.The measure- 3:50 ments showed that the LDi index accountsfor fast fluctuationsof sound levelbetter than other noiseindexes can do. In addition,the resultsindi- L6. Loudnessof sonic boom waveformswith atmosphere-induced catedthat thetime constant is a criticalparameter in measuringthe im- rounding.R. A. Sohnand D. W. Schowe(Douglas Aircraft Company, pulsecomponent of a sound.The soundrecordings were presented in the M/C CI-E41 (212-20),3855 Lakewood Boulevard, Long Beach, CA laboratoryat differentsound levels and in balancedorder to a groupof 90846) naivesubjects. They were asked to rateannoyance and impulsivity of each S31 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S31 soundstimulus using a ten-pointnumerical scale [G. Brambilla,M. R. Previoustheoretical investigations of activemeans to controlsound Carretti, and S. Santoboni,Proc. Noise 88, Vol. 3, p. 159, Stockholm radiationfrom a vibratingpanel have considered only the useof either (1988) ]. Correlationbetween the L m valuesand the subjectiveratings is acousticcontrol sources or vibration (point force) controlsources. As in progressin orderto testthe validity of thisindex in theevaluation of such,a generalunderstanding of the physical mechanism employed inthis both impulsivityand annoyanceof noise.[Work partiallysupported by noisecontrol technique has been lost. This paper presents a generalanaly- CEC. ] sisof the activecontrol of soundradiation from a simplysupported, baf- fledpanel. It considersthe useof both acousticand vibrationcontrol sources.The minimizationof the total soundpower output is usedas a criterionfor actuatorplacement, and minimization of thesound pressure at a pointis usedfor errormicrophone placement. Analytical results for minimizationof the soundpressure at a pointare verified experimentally. 4:35 It is shownanalytically that acousticcontrol sources operate by altering the radiationimpedance seen by the vibratingpanel. Vibration control L9. A generalanalysis of the active controlof soundradiation from a sourcesoperate by changingthe panelvelocity distribution, either de- baffled panel. Scott D. Snyder, Pan Jie, and Colin H. Hansen creasingthe amplitude of the offendingpanel mode, and/or alteringthe (Departmentof MechanicalEngineering, University of Adelaide,GPO temporalphasing of the panelmodes, thereby decreasing the effective Box 498, Adelaide,South Australia 5001, Australia) powerradiation from the panel. TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM H, 2:00 TO 4:00 P.M. SessionM. Underwater'Acoustics II: High-FrequencyAcoustic Imaging Nicholas P. Chotiros, Chairman alppliedResearch Laboratories, The Universityof Texasat dustin,P.O. Box8029, dustin, Texas 78713-8029 Chairman's Introduction--2:00 Invited Papers 2:05 M1. The STRESS sonar experiment. Darrell R. Jackson and J. George Dworski (Applied Physics Laboratory,College of Oceanand FisherySciences, University of Washington,Seattle, WA 98195) A high-frequencyscanning sonar experiment has been conducted as part of the STRESSProgram (sedi- menttransport events: slopes and shelves). An autonomousplatform was used to scana circularregion 100 m in radiuswith 5ø angular resolution. The acoustic frequency was 40 kHz andthe transmitted signal was an FM pulseof 2-kHzbandwidth. The objective was to quantifychanges in bottomscattering due to storm-induced sedimenttransport events. The acoustic experiment was conducted in waterof depth90 m at a siltysite off the Californiacoast. Other experiments measured environmental parameters such as sediment properties, bottom currents,and suspended sediment concentration. Scans were made at a rateof ten perday over a periodof 13 weeksduring the winter of 1988-1989.The acoustic data have been processed to determinescattering strength andthe correlationbetween scans as a functionof timeseparation. These results are presented in the formof imagesshowing the evolution of thescattering process during quiet and stormy periods. Very little changeis seenin the scatteringstrength, but the correlationmagnitude exhibits a monotonicdecrease that accelerated duringa storm.The phase of thecomplex correlation provides a very sensitive indicator of smallchanges. Two applicationsof thesephase data will bepresented. [Work supportedby ONR.] 2:25 M2. High-frequencyacoustic imaging of the seafloor.C. de Moustier(Marine PhysicalLaboratory, Scripps Institutionof Oceanography,A-005, La Jolla,CA 92093) For decades,sidescan sonars have beenthe primary tool to obtain acousticimages of the seafloor.Such imagesprovide qualitative information on the seafloorsurveyed based on amplitudevariations of the back- scatteredacoustic signals received. In the 1980s,bathymetric sidescan sonar systems, capable of simultaneous- ly producingacoustic images and measuring depth at numerouspoints across the swath, added a quantitative descriptionof theseafloor in theform of a depthcontour map. Similar claims can be made with multibeamecho sounderswell known for their high-resolutionswath bathymetry capabilities. Taking advantageof this high bathymetricresolution, the beamformedacoustic backscatter data can alsobe displayedas a geometrically correctacoustic image of the seafloorand providetextural informationnot availablein the contouredbathy- S32 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S32 roetryof the same area. Likewise, knowledge ofthe bathymetry, particularly bottom slopes, isneeded to correct forthe angular dependence of seafloor acoustic backscatter and construct a map of acousticbackscattering strengthover the area. Such a mapwill giveclues to regionalvariations in lithelogics. ContributedPapers 2:45 Acousticpusles entering a sandysediment through the water-sedi- mentinterface were observed. An acousticprojector located in thewater M3. Evaluation and verification of bottom acoustic reverberation columnwas usedto insonifyan array of sensorsembedded in the sedi- statisticspredicted by the point scatteringmodel. D. Alexandrou, ment.Carrier frequencies of from 5 to 80 kHz wereprojected toward the G. Haralabus(Department of ElectricalEngineering, Duke University, sedimentinterface at grazingangles below, near, and abovethe critical Durham, NC 27706), andC. de Mousticr(Marine PhysicalLaboratory, value. Sound-pressurelevels and time of arrival differencesassociated ScrippsInstitution of Oceanography,A-005, La Jolla,CA 92093) with the passageof the acousticpulse were determined from the sensor The pointscattering model offers a parametrizationof the reverbera- data. A weightedleast-squares algorithm was used to fit a planeto the tion envelopeprobability density function (pdf) in termsof the average arrivaltime differences yielding wave-front direction and speed estimates. numberof scattererscontributing to the return and the presenceof a Estimates of acoustic attenuation and wave-front coherence were also coherentcomponent in the receivedprocess. Computer simulations were obtained.[Work supportedby ONT underNORDA management.] usedto verifymodel predictions and to evaluatetheir usefulnessin the contextof seafloor classification. Central to the verificationstrategy was thesuccessful solution of theinverse problem based on synthetic reverber- ation data.The averagenumber of scattererswas determined from esti- 3:30 matesof the kurtosisof the instantaneousreverberation pdf. The magni- tude of a coherentcomponent embedded in the scatteredreturn was M6. Remotesensing of particle motionby crosscorrelation of acoustic recoveredfrom envelopehistograms with the assistanceof the Kolmo- backscatter.Nicholas P. Chotiros(Applied ResearchLaboratories, The gofer goodness-of-fittest. Following the verificationstudy, the modelwas Universityof Texasat Austin, Austin, TX 78713-8029} perturbedby introducingclustered and ordereddistributions of scatterers in addition to the standard Poisson. The initial results indicate that the In a previousdesign study, it wasdetermined that cloudsof passive reverberationenvelope pdf differssignificantly for the threescatterer dis- ridersin the upperocean, such as bubblesand suspendedmatter, could tributions.The clustereddistribution led to a rapidincrease in kurtosis, producedetectable and uniqueacoustic backscatter signatures. It is pos- whilethe ordereddistribution displayed evidence of intermittentcoherent tulated that the acousticsignatures may be processedto yield the move- scattering.The usefulnessof this parametrizationwas further tested with meritof the passiveriders, hence image the motionof the water itself. real reverberationdata representingseveral distinct seafloor regimes. Methodsbased on crosscorrelation processing were developed. Comput- [Researchfunded by ONR. ] er simulationsand experimentswere carried out to test the methods. [ Work supportedby ONR. ] 3:00 3:4,5 M4. Estimatian of sediment volume scattering strength arid absorl•tion coefficient.Robert A. Stewart and Nicholas P. Chotiros (Applied M7. Angular dependenceof seafloor acoustic backscatter at 12 kHz. ResearchLaboratories, The Universityof Texasat Austin, P.O. Box C. de Moustier (Marine PhysicalLaboratory, Scripps Institution of 8029, Austin, TX 78713-8029) Oceanography,A-005, La Jolla, CA 92093) and D. Aiexandrou (Departmentof ElectricalEngineering, Duke University,Durham, NC A simpletheoretical model for estimatingsediment volume scattering 27706) strengthand absorptioncoefficient from the reverberationtail of an acousticpulse observed by a pointreceiver within the oceansediment is The angulardependence of seaflooracoustic backscatter measured described.This modelassumes an infiniteplane wave entering the sedi- with a 12-kHz multi-narrow-beamecho sounderover a variety of sea- mentat normalincidence and scatteringisotropically from a uniformfield floors,ranging in depthfrom 500 to 5000 m, has beendetermined for of randominhomogeneifies. The modelpredicts the magnitudeand decay incidenceangles from 0 to 25 deg.The acousticdata consist of quadrature rate of the volumescattering at a point receiverin termsof the volume samplesof thebeamformed echoes received on eachof thesixteen 2.66- scatteringstrength and the absorptioncoefficient of the medium.By fit- degbeams era SeaBeam echo sounder. These data are subjected to adap- ting the predictionsof the modelto experimentallyobserved volume scat- live noisecanceling for sidelobeinterference rejection and to beam-to- teringdata, the volume scattering strengths and absorption coefficients of beam"decorrelation" to ensurethat the returnsare separatedin angle real sedimentswere inferred. Expelmentaldata were obtainedfrom in- with minimal crosstalk betweenadjacent beams. The measurementsare sedimentacoustic probes deployed at sitesoff Kings Bay, GA, and off then correctedfor geometriceffects, taking into accountthe ensonified PanamaCity, FL. The absorptioncoefficient results were compared with areafor eachbeam, which varies as a functionof the ship'sroll andbottom correspondingmeasurements from core samples.[Work supportedby slopes.It wasfound that apparentbottom slopes of lessthan 2 degin the ONT with NORDA management.] athwartshipsdirection are noticeable in thedata and must be compensat- ed for. For lack of systemcalibration, the data are presentedas relative meanenergy levels in l-degbins from 0- to about25-deg incidence. These resultsare comparedwith the resultsera REVGENcomputer simulation 3:15 whichfaithfully reproduced the acousticgeometry of the SeaBeam sys- tem and generatedscattered returns devoid of a specularcomponent. Re- M5. Plane-waveanalysis of acousticsignals in a sandysediment. Robert suitsare discussedbased on geometricand physicalarguments, and the A. Allenburg,Carl Faulkner,and Nick Chotiros(Applied Research abilityto differentiateseafloor types using such angular dependence func- Laboratories, The University of Texa.sat Austin, Austin, TX 78713} tions as assessed.[ Researchfunded by ONR. ] S33 J. Acoust.Sec. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of America S33 TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM C, 3:00 TO 5:05 P.M. SessionN. Architectural AcousticsII: Acousticsin Performing Spaces John W. Kopec, Chairman Rioerbank Acoustical Laboratories, 1512 Batavia A•enue, Geneva, Illinois 60134 Chairman's Introductions&00 ContributedPapers 3:05 a referencesource and the measuredpower emitted by the samesource placedon the stagesurface; mapping of isointensitylevels over the cou- NI. Is diffusionin concerthalls a propertyof the volumeor of the surface? pling area betweenstage and hall (prosceniumplane); evaluationof James B. Lee (Concert Acoustics, Box 18017, Portland, OR 972 i 8) acoustic-intensity-basedtransfer functions over a grid in the proscenium For nearly a century, the principalconcept of the theory of concert plane.A further descriptionof the theateracoustics was obtained by in- hallshas been that of a nearlyuniform field of energy,generated by many tensityvector measurements over three parallelgrids in the proseenium. reflections,that variesbul slowlywith time. This supposeduniformity Resultsof thisadditional characterization are presented as graphic views and isotropygave rise to the term "gooddiffusion," a propertyof the of acousticintensity flow lines. volume;it is difficultto quantify.An alternateapproach is to apply the photometricconcept of diffusereflection at the surfaces:This is readily quantified,permits calculation of intensityat the audience,and posits nothingabout the propertiesof the volume.Microcomputer programs to 4.'O5 implementthe conceptare at handand havebeen applied to a studyof BostonSymphony Hall, renownedfor itshighly diffuse ceiling and walls. N4. Verification of Bergeroh'sanalytleal methodapplied to room Modelingtechniques, also borrowed from photomerry,are availabletoo: acoustics.Hidemaro Shimoda (Institute of Technology,Shimizu 1/50th scalearchitectural models are eminently suitable; note that models Corporation,Etchujima 3-4-17 koto-ku, Tokyo i 35, Japan), employingmirrors---specular reflectors--are not accuraterepresenta- Norinobu ¾oshida, and Iehiro Fukai (Faculty of Engineering, tionsof goodconcert halls. HokkaidoUniversity, NI3 W8 Kita-ku,Sapporo 060, Japan) Applicabilityof the Bergeroh'smethod to roomacoustics has been studied,and was found useful for a transientanalysis in the timedomain [J. Acoust.Soc. Am. Suppl.I 84,S64 (] 988)]. In thispaper, a verifica- 3:25 tionstudy of theproposed analytical method is presented. Transient time responsesand reverberation times measured in a rectangularscale model N2. The organas a teachingresource in architectural acoustics,Bertram werecompared to the analyticalresults. The timeresponses at sixposi- Y. Kinzey, Jr. (Department of Architecture,University of Florida, tionsin theroom were observed by pulseexcitation with l/3-octave-band Gainesville, FL 32611 ) toneburst, and reverberation times were evaluated from these responses The studyof architecturalacoustics must include an acquisitionby usingSchroeder's integrated impulse method. Each time response at var- studentswith limitedor no musicalbackground of a rangeof knowledge iouspoints agreed very well with thatof thescale-model experiment and beginningwith the harmonicstructure of soundand its effecton timbre the averageof calculatedrevcberation times also agreed favorably with andcontinuing through an awarenessof theways, for betteror worse,the the experiment. architectural enclosure creates a reverberant field that affects various at- tributessuch as blend, clarity, and fullnessof tone and completesthe musicalsound that is heard.Aural experiencesare essentialif the student isto understandfully musicalsound and the roleof anarchitectural enclo- sure in its properaudition. The organ is an ideal instrumentfor demon- 4:25 strationsthat givethis experience.A seriesof live and recordeddemon- NS. Temporal changein interaural cross-correlationfunctions due to strationsusing the organis describedwhich have been used over several signal dynamics. H. Yanagawa (Ohmori R&D LabOratory, Pioneer yearsto showthe relationshipbetween music and architectural acoustics. ElectronicCorporation, i 5-5, 4-chome,Ohmori-Nishi, Ohta-ku, Tokyo 143, Japan) Maximal values of transient interaural cross-correlation functions (TRICC) for simpleburst signals were used earlier to analyzetemporal 3:45 changesin spatialimpressions such as soundimage localization and sub- N3. Application of vector intensity techniques for the acoustical jective diffuseness.These temporal changeswere found to correspond characterization of a baroque-style Italian theater. Domenico Stanzial, with subjectiveexperiences [Yanagawa et al., J. Acoust.Soc. Am. 84, E. Carletti, 1. Vecchi, and A. Fuschini (Ccmoter, National Research 1728-1733(1988) ]. for the presentstudy, a soundsignal was chosen that is closer to transient sources such as music or voice. It consists of two Councilof Italy, Via Canal Blanco28, 44044 Ferrara,Italy) burstsof whitenoise with a silentpause inbetwecn that isshorter than the Resultsconcerning acoustical measurements carried out in the Mu- reverberation time of the room. TR1CC values at the rise of the second nicipalTheater of Ferrara,built in 1790,are herepresented. Recent resto- whitenoise signal were analyzed to seehow theychanged with the length ration works replacedthe original woodenstage. A model, which de- of the pauseand with roomvolume. It wasfound that a shorterpause scribesthe enclosedspace as a retroactive linear system, has been causesthe fall (reverberation) of the initial soundburst to overlap the developedto characterizerooms acoustics. Modifications of the acoustic second,so Ihat TRICC values are smaller, but less so when the room is field,induced by the newstage, were measured by meansof the following smaller.These resultsclarify how temporalchanges in TRICC closely objectivetests: evaluation of the theatertransfer function having related reflectdifferences in roomswith regardto easeof soundimage localiza- inputand output, respectively, defined by the free-fiddacoustic power of tion. S34 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S34 4:45 spacetransfer functions. The all-passphase is associatedwith the rever- N6. Separationof total transferfunction phase into propagationand berantproperties of two-and three-dimensional spaces. The dereverbera- reverberantcomponents. Lan Liuand Richard H. Lyon(Department of lionof acousticalsignals in suchspaces can be accomplished by this sepa- MechanicalEngineering, Massachusetts Institute of Technology, ration. This paperdescribes the resultsof somerecent experimental Cambridge,MA 02139) studiesof soundpropagation in one-,two-, andthre•-dimensional spaces. A comparisonof the expectedand observedcomponents of reverberant Previouslyreported theory of phasedecomposition used the complex and propagationphase, with the resultsfrom theseexperiments, is pre- cepstrumto extractthe all-pass phase from the totalphase of acoustical sented. TUESDAY AFTERNOON, 28 NOVEMBER 1989 ST. LOUIS BALLROOM A, 3:00 TO 5:12 P.M. SessionO. SpeechCommunication III: FundamentalFrequency and Intonation GeorgeD. Allen, Chairman Departmentof Audiologyand SpeechScience, Purdue Unioersity, West Lafayette, Indiana 47907 ContributedPapers 3:00 arycan be found in theperception ofthe LH aswell. Synthesized nonsense words/mamama/wereprepared in sucha waythat both the onset of F0 Oi. Indivldual differencesin voice quality perception.Jody Kreiman, riseand the F0 peakoccur at variouslocations, while maintaining the BruceR. Gerratt, and Kristin Precoda(West Los AngelesVA Medical overallF0 contour(level-rise-peak-slight fall). The stimuliwere present- Center, Audiologyand SpeechPathology (126), Wilshire and Sawtelle ed to nativespeakers of Japaneseto determinethe boundarybetween the Boulevards,Los Angeles,CA 90073, and UCLA Schoolof Medicine, categoricalperception of LHH and LLH. The resultsshow that the LH CHS 62-132,UCLA, Los Angeles,CA 90024) sequenceis moreconstrained than the HL in termsof temporalalignment Sixteenlisteners judged the similarity of all possiblepairs of 18patho- of F0 changeand the syllableboundary. logicalvoices and, in a separatesession, 18 normal voices.Individual differences,multidimensional scaling was usedto derivea separateper- ceptualspace for eachlistener/voice set combination. These scaling solu- tionsaccounted for an averageof 83% of the variancein similarityratings for pathologicalvoices, and 77% for normalvoices. Listeners varied sub- stantially in the acousticcharacteristics they attendedto when judging 03. The frequencyscale of intonation.Dik J. Hermesand JoostC. van vocalsimilarity: Although all perceptualspaces included an F0 dimen- Gestel(Institute for PerceptionReseareh/IPO, P.O. Box 5 ! 3, N L 5600 sion,no other parameterwas usedby more than half the listeners,for MB Eindhoven, The Netherlands) either voiceset. Listenerswho sharedcommon perceptual dimensions Resultswill be presentedshowing that accent-lendingpitch move- often differed in the way they used the same acousticinformation. For mentsare perceived on a critical-bandscale. A sentencewas resynthesized example,F0 wasused as a continuousdimension by somelisteners, and to in two versionsdiffering in pitchand in formantfrequencies. The lower- sortvoices into groups(high- and low-pitchedgroups, pathological and pitchedversion sounded like a malevoice, the higherone like a female normal groups,etc.) by others;combinations of thesestrategies also oc- voice.One syllablewas rendered prominent by meansof a pitchmove- curred.Implications of theseresults for modelsof voicequality perception ment.The pitchcontours of the two versionsran parallelon anyof three will be discussed. frequencyscales: a logarithmicfrequency scale (semitones), a critical- bandscale, or a linearfrequency scale (Hz). In 2AFC experiments,sub- jectsindicated in whichversion the accentedsyllable was more promi- nent.Only whenthe excursionsof the pitchmovements were equal on a critical-bandscale, the choicewas random. In caseof equality in semi- 3:12 tones,subjects perceived the accent in the higherversion as more promi- 02. The perceptionof the low-high (LH) tonal sequence.Kazue Hata nent.In caseof equalityin Hz, the accentin the lowerversion was per- (SpeechTechnology Laboratory, 3888 StateStreet, Santa Barbara, CA ceivedas more prominent.These resultsallow a perceptuallymore 93105) and Yoko Hasegawa(Department of Linguistics,University of realisticmeasuring of the prominenceof accentedsyllables. [ Work sup- California, Berkeley,CA 94720) portedby !nstituulvoor Doven, St-Michielsgestel, The Netherlands.] It hasbeen reported that the primarycue for the HL tonalperception in Japaneseis not the actualF0 peaklocation but rathera fallingF0 contour.The F0 fall maybe significantly delayed, resulting in the F0 peak withinthe L-toned syllable. Furthermore, it wasfound earlier that ( 1) the 3:36 later the F0 fall in the L-tonedsyllable, the steeperthe fall rate required and (2) the fall mustbegin within the first two-thirdsof the durationof 04. Fundamental frequency and perceived prominenceof accented the vowel in the L-toned syllable.The presentexperiment investigates syllables.J. Terkenand R. Collier (Institutefor PerceptionResearch, whethera lackof synchronizationbetween F0 changeand syllablebound- P.O. Box 513, 5600MB Eindhoven, The Netherlands) S,3fi J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S35 In naturalspeech, accented words may differ as to thedegree of per- thesedata seem to indicatethat the spectral envelope is sufficientto syn- ceivedprominence. At theacoustic level, two aspectsof fundamentalfre- thesizethese voice qualities. Physiologically, it would appear that glottal quency(F0) variationmay be responsiblefor the perceivedprominence pulseshape is notsignificant. However, the spectral envelope may still differences:the magnitudeof F0 changesand the relativefrequencies of containcomponents of thesource such as subglottal coupling, spectral F0 maxima.Two experiments,with the samegroup of subjects,addressed slope,and/or the effectof the laryngealresonator on the vocalcavity thequestion of whichaspect of F0 betterpredicts perceived prominence. transferfunction. A demonstrationtape will be presented. Both experimentsused reiterant speech with synthesizedF0 contours. The speechmaterials consisted of'mamamamamamama' utterances with F0 maxima on the secondand penultimatesyllables ("PI" and "P2," respectively).In oneexperiment, subjects adjusted the frequency of P2 so 4:12 that it wasjudged to havethe samepitch as PI, for differentrates of baselinedeclination. In the secondexperiment, subjects adjusted P2 so 07. (Semi-)automaticpitch-synchronous computation of glottal flow. that it wasjudged to havethe sameprominence as P1, againfor different JacquesKoteman and Ben Cranen (PhoneticsInstitute, Nijmegen declinationrates. The resultsto be presentedare relevantfor refiningthe University,P.O. Box 9103, Nijmegen,The Netherlands) theory of pitch accentuation.For instance,if perceivedprominence is predictablefrom F0 maxima,both experiments should give the samere- Large-scaleinvestigations of the aerodynamicsof voiceproduction suits. are most easily performedusing noninvasivemeasurement methods. C}lottaiairflow can be estimatedby inversefiltering the mouthflow, mea- suredwith a pneumotachographmask. A semi-automaticpitch-synchro- nousinverse filtering method was developed that combinesthe advan- tagesof interactiveinverse filtering and inversefiltering with fixed-filter settings:Our methodyields accurate and calibratedestimates of glottal 3:48 flowduring VC and CV transitions,which are amenable for parametriza- O$. Contlnuafiveintonation in Mandarin. J. S. Mirza (Department of tion and canbe usedfor bulk processing.Using glottal dosing and open- Physics,National Universityof Singapore,Singapore 051 i, Republicof ing momentsderived from the electroglottogram,covariance LPC analy- Singapore) sis on the closedglottis interval is employedto compute the optimal inversefilter for eachperiod. For malespeakers, the methodyields quite The linguisticphenomenon of continuationin speechthat manifests stablewaveforms. These will beparametrized in orderto developrules for itself by modificationof pitch contour in a preconjunctionperiod was a physiologicallyinterpretable, time-varying synthetic voice source. The investigatedfor Mandarin.Mandarin is a tonallanguage; it hasfour tones. controlparameterg used are F0, duty cycle,top-top flowamplitude, glot- Its continuatireintonation was investigated by examiningselected parts tal leak area,and verticalphasing. A numberof theseparameters can be of intonationcontours prior to conjunctionsfor ninesentences each spok- derivedfrom boththe EGG and the glottalflow. Comparison of param- en by fivespeakers. Slopes of intonationcontours across the wordsimme- etervalues from the twosignals will givean indicationof their reliability. diatelybefore the conjunction were measured. It is reportedhere that the The results will be discussed. effectof continuationon Mandarin tonesis highly dependentupon the type of conjunctionsused. The continuationmostly drives up the slope valuesof Mandarin tones2, 3, and 4, while the slopevalue of tone I is pusheddown. The depression in slopevalue of tone1, however, cannot be takenseriously because its isolationmode value slope is closeto zeroand 4:24 the percentchanges can be erroneous.The continuationaffects on the averagetone 2 mostofall ( + 26%), thentone 3 ( + 21%), andthen tone 08. Fundamental frequency database with linguistic and phonetic 4 (4%). The conjunction"and" hasthe greatesteffect on the toneson the information.Masanobu Abe, Yoshinori Sagisaka{ATR Interpreting average( + 25%) followedby conjunction "but" (23%) and "and then" Telephony Research Laboratories,Sanpeidani lnuidani Seika-cho, (18%). Kyoto, 619-02 Japan), and Hisao Kuwabara (NHK Science and TechnicalResearch Laboratories, Japan) An importantproblem for speechscience is the relationshipbetween syntacticinformation, prosodic information, and fundamental frequency contour.To facilitatethe studyof the interactionamong these three fac- 4:00 tors, all three have been coordinatedin a continuousspeech database. Specificationsof the database are as follows. ( I ) Speechsamples consist of 06. Analysis and synthes's of six voice qualities. T. 503phoneme-balanced Japanese sentences spoken by a maleprofessional V. Ananthapadmanabha(AT&T Bell Laboratories,Murray Hill, NJ announcer[ Kuwabaraeta!., ICASSP '89, 560-563 (1989) I. (2) Phonet- 07974) and Jo Estill (Voice Consultant, 408 West 57 Street, New York, ic transcriptionsat severallevels of detailare provided[Takeda et al., NY 10019) Euro. Conf. SpeechTechnol. 2, 13-16 (1987}]. (3) Fundamentalfre- quencyis automaticallyextracted every 2.5 ms and extractionerrors are This is a preliminarystudy reportingon the acousticanalysis and correctedby hand. (4} The correspondingsentence is decomposedinto synthesisof six selectedvoice qualities; speech,falsetto, low-larynx, constituentwords and morphcrees with lexicalinformation such as inflec- twang,belting, and opera. The purposeof thisinvestigation is to testthe tionalcategories and is assigneda treestructure. This informationis se- abilityof the source-filtermodel to synthesizethis wide range of voice miautomaticallygenerated from input texts. (5) Each utteranceis seg- qualities.Five vowels,/i/,/e/,/a/,/o/, and/u/, in six voicequalities mentedinto minor phrases and each accent position is marked by listening spokenby a trained femalespeaker were recordedin a treatedsound to eachutterance. This fundamentalfrequency database has been used to booth, allowing SPL and F0 to free vary with a normal decay at the end of quantify fundamental frequencycontrol factorsand to show the effective- eachtoken. These recorded signals were digitized directly onto a comput- ness of this information. er. Inversefiltering of vowelsounds was performed using the linearpre- diction technique.Parameters of voice sourcewere extractedfrom the inversefiltered signal. Also, the spectralenvelopes of the vowelswere obtained.The sixvoice qualities were synthesized and compared with the originalrecording. Informal listeningtests indicated little differencebe- tween the original and the synthesizedtokens, thereby confirmingthe 4:36 analysisprocedure. To studythe relativecontributions ofsource and filter 09. Voeal fundamental frequency: Variation by language, language components,vowels were synthesized using the vocaltract transferfunc- group,and sex. Carolyn Wardrip-Fruin(Department of Communicative tion appropriatefor eachquality but usingan arbitraryvoice source. The Disorders,California StateUniversity, Long Beach,CA 90840) voicesource was varied in two ways;(i) keepingthe amplitudenatural butpulse shape fixed and (ii) keepingboth the amplitude and shape fixed. Vocalfundamental frequency was measured for speakersof fivelan- The differencesin voicequalities were still distinguishable.Pereeptua!ly, guagesunder three conditions (reading English, reading native language, S36 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society ofAmerica S36 and spontaneousspeaking-native language}. The sampleswere recorded muscles.Tetanic responsesof thesemuscles were normalizedand com- in a sound-treatedbooth and analyzed by a Vlsipitch(Kay Elemetrics) paredand resultsreported on tetanic90% contractionand 90% relaxa- frequencyanalyzer interfaced to an IBMxt computer.Preliminary analy- tion times.It was found that the thyroarytenoidlateral musclewas the sissuggests that meanfundamental frequency was surprisingly similar fastestmuscle in thegroup with the meantwitch contraction time of 12.3 acrosslanguages for the variousspeaking conditions, and that themean ms and meanhalf relaxationtime of 11.9 ms,and thyroarytenoidroedial fundamentalwas higher for readingthan for speaking(as hasbeen found muscle had mean twitch contraction time of 21.2 ms and mean half relaxa- in studiesof English),but that therewere significant differences between tiontime of 18.6ms. [ Work supportedby N INCDS Grant No. NS 16320- languagesand by sexin standarddeviation of the fundamentalunder the 07.] variousspeaking conditions (reading English, etc. }. The resultssuggest that fundamentalfrequency is determinedprimarily by physiologicalfac- torswith somelinguistic variations. 5:00 OIL Changes in vocal fold length with nerve stimulation in canine 4:48 larynges.Jiaqi Jiang and Iago R. Tilze (Department of Speech Pathologyand Audiology,University of Iowa, Iowa City, IA 52242} O10. On the mechanical prol•erties of laryngeal muscles. Fariborz Allpour and Abdolali Najafi (Voice Acoustics and Vocalfold length change and adductory movement of theglottis were BiomeehanicsLaboratory, Department of Speech Pathology and obtainedfrom sevencanine larynges when the recurrentand the superior Audiology,The Universityof iowa, Iowa City, IA 52242) laryngealnerve were stimulatedbilaterally. A surgicalprocedure was usedto makemarks on the vocalfold for lengthmeasurements. Frame by Active propertiesof canine laryngeal muscleswere investigated frame analysiswas usedto get staticand dynamicdata on vocal fold througha seriesof experimentsconducted in vitro.Samples of the thy- length.The meanmaximum elongation of the membranousvocal folds roarytenoidroedial and lateralmuscles, also, cricothyroid pars recta and was44.7% with 120-mselongation time and 140-msrelaxation time. The pars oblique, were dissectedfrom dog laryngesexcised a few minutes meanmaximum shortening of membranousvocal folds was 17.8%, with beforedeath and kept in Krebssolution at a temperatureof 37 ñ i øCand 110-mscontraction time and 1i 5-msrelaxation time. When both agohist a pH of 7.4 ñ 0.05. Field stimulationwith parallel-plateplatinum dec- and anlagonistmuscles were in maximumcontraction, the meanelonga- trodeswas applied to studytwitch and tetanicresponses of thesemuscles tion was 24.3%. The length-time curve approximatesan exponential in isometricconditions. The activeforce of each samplewas recorded function.Closing of glottisoccurred in lessthan 35 ms.The temporal electronicallywith a dual servosystem (ergometer). Resultsare reported resultsagree with earlierpitch change results. [Work supportedby NIH on the twitch contractiontimes and half relaxationtimes of laryngeal Grant NS16320. ] S37 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of America S37 WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM F, 8:00 A.M. TO 12:00 NOON SessionP. Noise II: Measurement of Employee Noise Exposure Richard M. Guernsey,Chairman R. M. Guernsey& Associates,P.O. Box ]517, Morristown,New Jersey 07960-]517 Chairman's Introduction---8:00 Invited Papers 8:05 P1. Usingnoise dosimeters in anindustrial noise control program. John œrdreich(Ostergaard Associates, 115 Bloomfield Avenue, Caldwell, NJ 07006) Traditionaluse of thenoise dosimeter to assessemployee exposure is to satisfya regulatorymandate. Issues ofeffects of microphone placement and statistical reliability of the measurements have limited the usefulness of exposuremeasures soobtained and furthermore, have deflected attention from the real utility of thedevice to providecritical information for noisereduction. Experience with measurementson maintenanceworkers showsthat time histories that are provided by certain dosimeters allow the engineer to identify noise generating processesthat produce the major components of the exposure. As a practicalmatter, the high-level processes thatcontribute several decibels to thetime-weighted-average exposure are of greaterconcern than statistical variabilityinduced by other factors. For thisreason, acquisition of timehistories in employeeexposure assess- ment is encouraged. 8:30 P2. Monitoring noise exposureusing employeemovement analysis techniques. Thomas $. Bragg (Noise Unlimited, Inc., 9 Saddle Road, Cedar Knolls, NJ 07927) In mostindustrial activity, employee work routinesvary from oneday to anotheror from weekto week. TheUnited States Occupational Safety and Health Administration (OSHA), recognizingthis fact, has speci- fiedthat an employee'slong-term noise exposure should be expressed as a summationof partialexposures to varyingsound levels. A techniquehas been developed that analyzes employee movement, breaking an average dayinto a seriesof locationsor activities,each with an associatedtime duration. Sound level data are then acquiredfor eachlocation or activity.A computeris then used to calculatethe partial exposure, as defined by anycriterion, for eachlocation or activityand thesearc summedto expressthe overallexposure level for each employeejob function.The computer also generates a list of noisecontrol priorities and the minimum amount of noisereduction required at eachlocation to achievecompliance with the selectedcriterion. Maximum priorityis given to thelocation which, when attenuated, will givemost benefit to mostemployees. Determining theseminimum reduction levels is mostimportant, as the costand feasibilityof noisecontrol arc directly relatedto the amountof reductionrequired. 8:55 P3. Asystemtomonitorthenoiseexposureofpersonswearingheadsets.W.K. VanMoorhemandK. S. Woo (MechanicalEngineering Department, University of Utah, SaltLake City, UT 84112) Althoughlegislation limiting noise exposure in thework place applies to individualswho wear headsets as partof theiremployment, no methodfor measuring the noise exposure of theseindividuals has previously existed.A systemfor carrying out this measurement has been developed. This system is indirect,in thatthe electricalsignal to theheadset ismeasured, the transfer function of theheadset, ear, and dosimeter microphone isapplied to the measured signal, and the resulting electrical signal input to a dosimeter.This approach requires the transfer function of the headsetbe measuredfor each brand and model that is used,but eliminates the need forin-ear microphones of the more direct approach. The transfer function of the headset ismeasured using an acousticmanikin head.A limited numberof headsetsbrands and modelshave beenused with this systemto investigateboth the variations between models and within a model.[Work supported by Departmentof Labor. ] $38 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America 9:20 P4. Probemicrophone techniques for determiningexposure to noise.Mead C. Killion (EtymoticResearch, 6 I Martin Lane,Elk GroveVillage, IL 60007) Traditionaldosimctcr measures, using a micropboncin tbesound field, provide litde information about the actualnoise exposure experienced by a subjectwearing carplugs, carmuffs, earphones (traditional or insert),or someoneexposed to a highlydirectional or localnoise source. In theseinstances, the noiseat the eardrum wouldappear to bea moreuseful measure, but cannotbe directly compared to traditionaldamage risk criteria becausethe latter are basedon measurementsin a diffusesound field. If a simplebridged-T filter network [Killion, J. Audio Eng.Soc. 27, 13-16 (1079) ] correctingfor the transferfunction between the diffuse and the eardrumis interposedbetween a fiat-frequency-responseprobe microphone and a dostmeter,the resultingdosimeter indication can be compareddirectly to traditionaldamage risk criteria.Techniques for minimizingthe potentialerrors in suchmeasurements will bediscussed. 9:45 P5. Measurement of hearing protector attenuation using the physical ear attenuation test (PEAT). Kevin Michael (Departmentof CommunicationDisorders, The PennsylvaniaState University, 3F Moore Building, University Park, PA 16802) The attenuationprovided by a sampleof circumauralhearing protectors (ear muffs) wasmeasured using thephysical ear attenuation test (PEAT). The PEAT utilizeda smallelectret microphone mounted in a silicon earplug in theouter ear of a humansubject. The subjectswere exposed to a highlevel {about 100 dBA) of pink noisewith andwithout the hearing protector fitted. The attenuationprovided by the protectorwas calculated by subtractingthe spectrum measured under the protector from thespectrum measured without the protector. The resultsof the PEAT evaluationwere compared to the resultsobtained using the testmethod described in ANSI StandardS12.6-1984, the Method for the Measurementof the Real-EarAttenuation of HearingProtec- tors.Agreement was found between the methodsfor the testfrequencies from 250 to 8000Hz. The PEAT results at 125 Hz were lower than the results obtained with the ANSI S12.6-1984 method. 10:10-10:30 Break ContributedPapers 10:30 of a sourcecan be measured, sound power determined by integrationover the radiationarea, and the sound-pressurelevels at a distancefrom the P6. Comparisonof ANSI noise dosimetersand IEC personalsound sourcecan he calculated. An assumptionismade when the sound power to exposuremeters. Robert Krug (Cirrus Research,Inc., 6818 W. State sound-pressurecalculation is made,that at low frequencies(20-100 Hz ) Street, Suite 170, Wauwatosa, Wl 53213) thewaves have propagated and the resultant sound-pressure level is from ANSI and IEC haveprepared draft standardson metersto measure the source.This work teststhe reliabilityof that assumption.A speaker workerssound exposure. The standardshave a number of similarities. wassuspended inside three different size boxes, and poweredby a noise Bothare based on IEC 804,integrating-averaging sound level meters, and generatorbandpassed at discretefrequencies. Sound intensity measure- both are intendedto specifymeters worn on a workerand to measure mentswere recorded near field at the samediscrete frequency. Free-field noisethroughout a work day. The ANSI draft specifiesmeters with ex- soundpressure measurements were then recorded at distancesfrom the changerates of 3, 4, or 5 dB, with a fast or slow time constantand a boxes.This paperreports the resultsof the measurements. threshold.The IEC draftspecifies only a 3-dBexchange rate with notime constantor threshold.The ANSI draft specifiesa read out in percentdose whilethe IEC draft is developedaround Pascal squared hours. For con- tinuousnoise above the threshold,the readingcan be converted from one setof unitsto the other.For intermittentand impulsive, a noisesignificant 11:00 difference will occur. PS. Reverberationand noise propagation measurements in a powerplant. Frank H. Brittain and Ralph W. Passage(Bechtel Corporation, P.O. Box 3965, San Francisco,CA 94119) The acousticalcharacteristics of emptyspaces are relativelywell 10:45 known.An understandingof roomacoustics is important both for design- P7. Low-frequencysound intensity measurementand free-field sound- ing spacesand for controllingnoise from equipment in enclosedspaces. Power plantsand similar industrialplants pose a challengebecause their pressure propagation. William D. Gallagher and Lawrence spacesare not only very large, but also "full" of equipment.These spaces F. Heitkamp(Engineering Specialties Corporation, 8147 Delmar, Suite 207, St. Louis, MO 63130) containa high concentrationof rotatingequipment, piping, structural steel,cable trays, gratings, and tanks.Measurements of reverberation Classicalacoustical equations relate the soundpower of a sourceto timeand noise propagation were made in a powerplant. The equipment in sound-pressurelevel at a givendistance from that source. Sound intensity thesespaces changes the acousticalcharacteristics of the spaces.These $30 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S39 measurementsshow that soundabsorption properties of the spacesare Engineering,NCSU, Raleigh, NC 27695-7910),and Richard G. Pearson greaterthan are expected from published sound absorption data and theo- (Departmentof IndustrialEngineering, NCSU, Raleigh,NC 27695- riesof roomacoustics. Implications of the resultsfor noisecontrol of 7906) powerand other industrial plants are discussed. A hearingconservation program (HCP), originallydesigned for an 8- h workshiftschedule, was evaluated at a plantsite where a 12-hworkshift scheduleis now utilized. The studyincluded three phases: a noiseanalysis of the work environment,HCP evaluationusing audiometric database 11:15 analysis(ADBA), and an evaluationof the hearingprotection devices (HPDs) usedat the facility by measuringtemporary threshold shift P9. The aircraft noise-inducedpermanent threshold shift. 'Wu Yong- (TTS) andapplying ADBA procedures.The overall measured employee xiang, Wang Bing-guang,and Wang Xin-yu (Instituteof Aviation TWA in the processarea where the TTS studywas conducted was 92 Medicine,Air Force,PLA, Xi Diao Yu Tai 30 (East), Beijing,People's dBA. The ADBA resultsindicated that the existingHCP is, at best,mar- Republicof China) ginal.The mostlikely causes of thisless than desirablerating are inade- quateaudiometric testing procedures and inadequate HPD utilization.It The hearingof 410 groundcrewwho had workedunder 100-130 was concluded that the introduction of the 12-h workshift has had no dB(A) aircraftnoise was measured. The groundcrewwere divided into impacton thepresent effectiveness of the HCP. All threeHPDs in useat four groupsby the lengthof service:less than 5 years,5-9 years,10-14 the facility (3-M foamearplug, E-A-R foamearplug, and BilsomSoft years,and 15-20 years.Their audiogramswere taken with a Kamplex earplug)offered effective protection from noiseat all audiometrictest audiometerin a soundproofmobile cabin, in whichbackground noise was frequencies(0.5 to 6 kHz) except0.5 kHz. All threeHPD wearergroups 23-25 dB(A). One-hundredand forty sevenmilitary menof comparable exhibitedmean TTS at 0.5 kHz, which was significantlydifferent from agewho had not been exposed to excessive noise were selected as a control zero at the 0.05 levelfor the E-A-R and 3-M wearergroups. group.Their audiograms were taken in thesame way. The results indicat- ed that the hearinginjury was an actual menaceto the groundcrew. Amongthem the high-frequency hearing loss and the noise deafness made up46.1% and6.1%, respectively.Compared with the control group of the sameage, the differencesof the hearingaverage value of eachfrequency weresignificant. Their hearinginjury was aggravated with the lengthof service.The high-frequencyhearing losses of the fourgroups, less than 5 11:45 years,5-9 years, 10-14 years,and 15-20 years,were 32.7%, 47.3%, Pll. An improvedmodeling technique for high-orderacoustic spectra. 54.4%, and 67.3%, respectively.In orderto preventhearing damage, the groundcrewought to wearear protectorsand the groundnoise suppres- Vijay Raman, Osama Mohammed, Kang Yen, and Kurt Ramdin sorsshould be usedwhen the aircraft engineis operating.It was discov- (Departmentof ElectricalEngineering, Florida International University, Miami, FL 33199) eredthat the "V"-shapeddepression was not at 4 kHz but at 6 kHz in 410 groundcrew.There were similar phenomena in domesticand foreign lit- The analysisand modelingof noisyenvironments are consideredin eratures.The formerviews concerning the pathogenesisof the "4-kHz thiswork. Parametric modeling is often the mostconvenient, since it can depression,"therefore, is open to question.This is a quiteinteresting sub- be usedas a basisfor parametricfiltering and adaptive filtering for noise ject that warrantsfurther study. reduction.The earlierwork of thes.eresearchers dealt with the develop- mentof autoregressive(AR) as well as the autoregressive-moving-aver- age(ARMA) modelingof a dataset taken from digitized recordings of a typicalindustrial environment. It wasshown therein that the order of the modelrequired is extremelyhigh due to the periodiccomponents, and is 11:30 thereforenot asuseful as expected. In thiswork, a newapproach is devel- opedto modelperiodic components and the random part sequentially, not P10. Hearing conservationprograms (HCPs): The effectivenessof one separately.This method is able to model the data set with much fewer company's HCP in a 12-h workshift environment.Jacqueline coefficientsthan in the standardparametric models, and is usefulin de- L. Reynolds(Department of Industrial Engineering,NCSU, Raleigh, signingthe noise-reductionfilter. [Work supportedby the FloridaHigh NC 27695-7906), Larry H. Royster (Department of Mechanical Technologyand Industry Council.] WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM G, 8:00 A.M. TO 12:00 NOON SessionQ. Physical AcousticsIII and Bioresponseto Vibration II: Biophysical Mechanisms as Bases of Safety Criteria for Medical Ultrasound II Edwin L. Carstensen,Chairman RochesterCenter for BiomedicalUltrasound, University of Rochester,Rochester, New York 14627 ContributedPapers 8:00 Engineering,University of Michigan Medical Center, Ann Arbor, MI 48109-0553) Q1. Observation of acoustic cavitation in excised canine urinary bladders.J. B. Fowlkes,P. L. Carson,E. H. Chiang, and J. M. Rubin A high-intensityultrasonic field has beenused to generatebubbles (Department of Radiology, Division of Radiological Physics and within urinarybladders excised from dogs.Following the exposure,bub- S40 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S40 bleswere easily visualized using an ATL UltraMark9 diagnosticscanner frequenciesbetween 0.25 and 1.68 MHz. Thetemperature elevation at the witha 5-MHzmechanical sector scanhead. Scattering ofthe high-intensi- focalzone at thedepth of 5 cmin thetissues was measured during the ty ultrasoundby the bubbles was also observed asuniformly high ampli- soundpulse using a twistedpair thermocouple (wire diameter 50/•m ). A tudescan lines during the irradiation. The bladders were removed surgi- focusedhydrophone was also used to sensethe subharmonicemission callyafter tying off the uretersand urethrato preventurine loss and fromthe focal zone. At eachthermocouple location the pulses were re- exposureto externalcontaminants. Each bladder was sealed in a plastic peatedwith increasing applied acoustic power until the temperature ele- bagfilled with a degassedaqueous solution of sodiumchloride (0.9% by vationbecame irregular with sudden temperature peaks during the l-s weight)and sodium azide ( 1 g/liter). A bagcontaining a bladderwas Pulseand with an appearanceof subharmonicemission of noise.Similar centeredin a sealeddegassed water path at thecommon focal point of a 7- eventswere also observed at higherintensities during low-duty cycle cm-diam, 500-kHz transducer, and a 10-cm-diam brass reflector. The pulses,although not with pulses including only I or 2 cycles.These results transducerand reflectorwere both focused at 10 cm and werealigned indicatethat transient cavitation can appear at intensitiesthat could be coaxially.Exposures were 10 s in durationbut intensity or pressurevalues reachedduring clinical scanned focused ultrasound hyperthermia (400- werenot availableas of the writingof thisabstract. No significantdisco- 500Wcn 2at 1MHz) andthus can significantly modify the temperature lorationof theexterior bladder wall was observed within the beam path. distributionand the biological effects of thetreatment. [Work supported Resultsof this study suggest that bubbles might be externally generated in by NCI Grant No. CA 46627.] vivofor useas echocontrast agents and for studyingthe bioeffectsof acoustic cavitation. 9:00 8:15 QS. Comparisonof levelsfor ultrasoniceavitation in neonatalmice and other mammals.Leon A. Frizzell (Department of Electrical and Q2. Excitationof surfacemode activity of bubblesin ultrasonicfields. ComputerEngineering, University of Illinois, 1406 West Green Street, Robert K. Gould (Departmentof Physics,Middlebury College, Urbana, IL 61801 ) Middlebury, VT 05753) The levelfor cavitationinvolvement in hind limb paralysisof the Whena gaseousbubble finds itself in a liquidin whichan ultrasonic mouse neonate has been determined at 10 øC and 37 øC at 1 MHz. These fieldexists, the bubble may undergo radial, orderly surface mode or chao- levelsare compared to levelsfor cavitationas reported in othermamma- ticsurface oscillations depending on the size of the bubble, strength of the liansystems. Results have also been obtained for pulsed exposures, and a soundfield, and other parameters such as viscosity of theliquid and fre- comparisonis madebetween results from continuouswave and pulsed quencyof thesound. In theevent of purelyradial oscillations there is little exposures.It is suggestedthat the pulseaverage intensity and total sound acousticstreaming in the vicinityof the bubble.However, when bubble on timemay be more useful parameters (as opposed to timeaverage inten- surfaceactivity is present, vigorous streaming will occurnear the bubble. sity and exposureduration) for comparingpulsed and continuouswave Experimentalstudies of a bubblelevitated in a 20-kHz standing-wave cavitationresults within certain exposure regimes. This may prove to bea soundfield in a chamberof waterhave included the obtaining of strobo- usefulapproach to theincorporation of continuouswave results, relating scopicmovies of bubblesundergoing orderly surface modal vibrations of to cavitation,in the determinationof safelevels of exposureto diagnostic orders2, 3,and 4, therebeing no existence of mode 1. The frequency of the ultrasound. surface oscillations is half that of the sound field. Movies of modal oscilla- tions 2, 3, and 4 will be shown. Movies of mode 4 were obtained for the firsttime this year by George E. Lawrence,while carrying out his senior thesis research. 9:15 Q6. Biophysicalmechanisms in the formulationof Canadianguidelines for diagnosticultrasound devices. Stephen H. P. Bly andD. A. Morison 8:30 (Department of National Health and Welfare, Health Protection Building, Room 66, Tunney's Pasture, Ottawa, Ontario KIA 0L2, Q3. Radial oscillationsof gasbubbles in viseoelastiematerials. Charles Canada) C. Churchand Ronald A. Roy (NationalCenter for PhysicalAcoustics, University,MS 38677) A mechanisticassessment of ultrasound safety is presentedas a meth- od to formulateguidelines for diagnosticultrasound devices. This ap- Mostprevious studies ofacoustic cavitation inbiological tissue have proachwas adopted since surveys had indicatedthat ultrasoundoutput treatedthe materialsurrounding the bubbleas a Newtonianfluid with levelsfrom someexisting devices were higher than the no-effectlevels sometimesincreased viscosity. The presentwork describesthe derivation found in epidemiologicalstudies and studiesof mammalianbiocffects. of a generalizedRayleigh-Plesset equation for a gasbubble in a contin- Existingstudies on twomechanisms, heating and cavitation, were exam- uousviscoelastic material. Viscoelasticity is modeledto includeeither ined.It wasfound that knowledgeof the cavitationmechanisms in vitro stressrelaxation (Maxwell fluid) or aftereffect(Voigt fluid). Several andin nonmammalianmodel systems could not be extrapolated to predict casesare considered: ( 1 ) a bubblewith a (a) viscousshell suspended in a cavitationthresholds in humans.For the thermal mechanism,sufficient Newtonianfluid, (b) viscoelastieshell suspended in a Newtonianfluid, data were available [ AIUM BioeffectsCommittee, J. Ultrasound Med. 7, (c) viscoelasticshell suspended in a viscoelasticfluid; and (2) a "clean" S1-S38 (1988)] to formulatea guideline.The guidelinerecommends bubblesuspended in a (a) viscousor (b) viscoelasticfluid. In general,the that, for diagnosticultrasound devices used for fetal examinationsand effectof viscoelasticityisto decrease the resonance frequency and increase yieldingfree-field spatial peak temporal average intensities greater than thethreshold for transientcavitation. Preliminary measurements of cavi- 500mW/cm 2, estimates should be provided of maximum local tempera- tationthresholds in a liquidseeded with Albunex© stabilizedmicrobub- ture elevations in fetal tissues. blesare presented and interpreted in lightof thetheory. [ Worksupported by NIH. ] 9:30 8:45 Q7. Intensitygain criteria for focusedultrasound tumor hyperthermia: Theory and in vitro experiments.Brian J. Davis and PadmakarP. Lele Q4. The thresholdof thermally significantcavitation in musclein oioo. (Department of MechanicalEngineering, Massachusetts Institute of Kullcrvo Hynynen(Department of RadiationOneology, University of Technology,Cambridge, MA 02139) Arizona, Health SciencesCenter, Tucson, AZ 85724) Plano-concavelenses and concave radiators used in scanningfocused Thetransient cavitation threshold was measured by dogs'thigh mus- ultrasoundsystems have well-established capabilities for deepand selec- clesin vivoduring a l-s exposureto continuousor pulsedultrasound at tive tumorhyperthermia. The focalintensity gain, a goodpredictor of S41 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S41 selectiveheating, is expressedas a functionof the treatmentdepth below a heatingwith "pivoted" and circular scanned focused beams are presented. planarbeam portal integument,frequency-dependent tissue absorption Measuredintensity fields are comparedwith numericalmodels that aim rate, wavenumber,and apertureangle. From this expressionnovel and to ( 1) solvethe diffraction pattern of an absorbingspherical lens coupled usefulrelations are derivedfor the maximumfocal gain as a functionof to a clampedpiston radiator, (2) predictthe intensityfield of a focused systemspecific maximum aperture angle, optimum frequency, and other beamin an absorbingmedium, and (3) accountfor lenspropagation that parameters.Temperature distributions from in vitroand tissuephantom is not parallelto the beamaxis. [Work supportedby USPHS Grant No. experimentsdesigned to establishminimum gain criteria for selective 31303 and the Lutheran Brotherhood.] 9:4•-10:00 Bre•Lk 10:00 (1958) ]. Usinga lightscattering technique, the threshold for inceptionof individualsurface waves was examinedand the resultswere compared QS. Estimationof meanscatterer size from sparse,random distributions with both analytical and numericalcalculations of this threshold.The of polystyrenespheres in solutionvia Mie scatteringand cepstral signal observedand calculatedvalues are generallyin agreementalthough the processing: Theory and experimental results. C. R. Meyer, J. identityof the surfacewave mode is difficultto determine.[Work sup- B. Fowlkes,P. L. Carson, H. Wang, and H. Chan (Department of portedin part by the Officeof Naval Research.] Radiology,Division of RadiologicalPhysics and Engineering,University of MichiganHospitals, Ann Arbor, MI 48109-0553) Four solutionsof narrowlydistributed polystyrene spheres of onedi- ameter,either 41, 136,271, or 381•um, sparsely suspended in degasseal 10:30 water, wcrc imagedusing multiple 5- and 7.5-MHz medicalultrasonic transducers.Radio-frequency backscattering data from the 136-, 271-, Q10. The applicationof Hamilton's principle to a bubbly liquid. J. and381-pro solutions were processed using the data from the 41-/•m solu- A. Hawkins,Jr. (Applied ResearchLaboratories, The Universityof tion as a volumetric reference to correct for transducer diffraction effects Texasat Austin,Austin, TX 78713) and A. Bedford(Department of andacquisition system gain vibrations. The powercepstrum of themean AerospaceEngineering and EngineeringMechanics, The Universityof logdifference spectra averaged over all overlappingdata segments was Texas at Austin, Austin, TX 78712) computedand corrected for theeffects of attenuationand Rayleigh scat- tcringin thereference. Abcissas of thecepstral peaks were plotted versus Naturallyoccurring bubble populations have a distributionof bubble scattererdiameter. The zero-intercept, linear regression line for thelong- sizes.Well-established continuum theories exist for the propagationof er-lagpeaks yielded a slopeof 3.4/•s/mm,r• = 0.997,corresponding to a acousticwaves through a bubblyliquid with a singlebubble size. The Mie-region periodicityin backscatteredcross section of ]ca= 0.6. This applicationof thesetheories to bubblyliquids with severalsizes is neces- valueis in excellentagreement with calculationsbased on thescattering sarilylimited. Here two approachesare describedbased on Hamilton's theoryof Faran,which includes shear waves and yields a periodicityof principlefor analyzingthe acousticsof bubblyliquids containing a distri- ]ca = 0.66. butionof bubblesizes. First, the mixture is treated as a liquidcontaining a bubble"continua" of severaldiscrete sizes. The analysisleads to N + 2 (N is the numberof differentbubble sizes) simultaneous equations in the motion,density, and volume fraction, which must be solved numerically. The systemwas solved for one, two, three,and fivedifferent bubble sizes. 10:15 The secondapproach is to assumethat a continuousdistribution of bubble Q9. Thresholdsfor surfacewave generation on acousticallylevitated gas sizesexists in the liquid. Surprisingly,the secondapproach leads to a bubbles. Steve Hotsburgh, R. Glynn Holt, and Lawrence A. Crum relativelysimple expression for the wavenumber.It wasshown that the (National Center for PhysicalAcoustics, University of Mississippi, phasevelocity and the attenuationof acousticwaves can be determined University,MS 38677) for an arbitrarydistribution of bubblesizes. The two approachesare shownto agreein the limitingcase of discretebubble sizes and to agree When a gasbubble is acousticallylevitated in a standingwave, the with the resultobtained by Commanderand Prosperetti[ K. W. Com- sphericallysymmetric volume oscillations can readily couple into surface manderand A. Prosperertl,J. Acoust.$oc. Am. 85, 732-746 (1989)]. waves[M. Strasbergand T. B. Benjamin,J. Acoust.Soc. Am. 30, 697 [Work supportedby ONR, IR&D Programof ARL:UT.] S42 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S42 10:4S-12:00 Panel Discussion Applicationof Informationon Mechanisms in theDevelopment of ExposureCriteria PANEL CHAIRMAN: A. R. Williams Universityof Manchester Departmentof Medical Biophysics ManchesterM13 9PT, United Kingdom PANEL MEMBERS: StephenH. P. Bly Bureau of Radiation and Medical Devices Room 233, Environmental Health Centre Ottawa, Ontario K1A OL2, Canada Mel E. Stratmeyer Centerfor Devicesand RadiologicalHealth 12709 TwinbrookParkway (HFZ-112) Rockville,Maryland 20857 Gail ter Haar Institute of CancerResearch Clifton Avenue Sutton,Surrey SM2 5PX, UnitedKingdom Marvin C. Ziskin Department of DiagnosticImaging Temple UniversitySchool of Medicine Philadelphia,Pennsylvania 19140 WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM B, 8:00 A.M. TO 12:00 NOON SessionR. PhysiologicalAcoustics I and Psychological Acoustics III: Ear Canalto Cortex:Quantifying Auditory Mechanisms Julius L. Goldstein, Chairman CentralInstitute for the Deaf, 818 SouthEuclid Avenue, St. Louis,Missouri 63110 ContributedPapers 8:00 use of bandlimited noise as a stimulus results in decreased coherence betweenthe two microphonesignals relative to the coherenceobtained R1. Measurementof the local powerreflection coefficient in the ear canal with sinusoidalor otherdeterministic signals (e.g., a chirpsignal ) at high using the two-microphonemethod. W. J. Murphy, A. Tubis frequencies.The effectson the coherenceand the reflectancedue to sinu- (Departmentof Physics,Purdue University, West Lafayette, IN 47907), soidalchirp and bandlimitedstimuli are presented.The measurements andG.R. Long(Department of Audiologyand Speech Sciences, Purdue wereperformed in an acrylichard-terminated tube. As a testof the as- University,West Lafayette,IN 47907) sumptionof local onedimensionality of the ear canal,the reflectancewas measuredas a functionof the insertiondepth of the two-microphone The local power reflection coefficientin the ear canals of humans has probe in several subjects.At high frequencies,the measured reflectance beenmeasured using the two-microphonetechnique [W. J. Murphy,A. wasfound to varysignificantly in earcanals with contortedgeometries. In Tubis,and G. R. Long,J. Acoust.Soc. Am. Suppl.1 82, S70( 1987)]. The relativelystraight ear canals,the measuredreflectance was found to vary accuracyof the calibrationtechnique and the assessmentof the random by lessthan 10% for changesof 5 mm in the insertiondepth. [Work andbias errors associated with the reflectance have been investigated. The supportedby a grant from the DeafnessResearch Foundation.] $43 d. Acoust.Soc. Am.$uppl. 1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S43 8:15 less(8.1 dB) energywas required for elicitationof thresholdresponse in the shallowposition as comparedto the two other.However, no statisti- R2. On the passive power absorption potential of the eardrum. R. cally significantdifferences were found betweenthe most appropriate D. Rabbitt (Department of Mechanical Engineering, Washington (full and intermediate)insertion depths at any testedfrequency. Al- University,Campus Box 1185,St. Louis, MO 63130) thoughREAT valuesshowed increased attenuation for the frequencies Experimentalevidence reported by Khanna[ Int. Cong.Acoust. B6- 0.125-1 kHz asthe eartipwas inserted more deeply in the earcanal, at all 4, Toronto(!986) ] andAllen [ PeripheralAuditory Mechanisms (Spring- threeinsertion depths over all testedfrequencies, values were more than er, New York, 1985), pp. 46-51 ] suggeststhat a verylarge fraction of the 15 dB greaterthan the attenuationobtained with the conventionalear- incidentacoustic power in the cat ear canalis absorbedat the eardrum.In phone. someanimals, the high efficiencyextends over a very wide frequency rangefrom 300 Hz-30 kHz. Thesefindings are sensitive to the condition of the animal and are subjectto relativelylarge interanimalvariations, particularlyat highfrequencies. Theoretical models describing the func- tion of the middleear differsignificantly in predictingthe powerabsorp- tion andin predictingthe fractionof the energyflux dissipatedwithin the 9:00 eardrumitself. Someinsight into the behaviorof the eardrumcan be gainedby studyingthe powerabsorption properties when the ossicular RS. Distortionproduct emissions as predictorsof frequency-specific chainis immobilizedor removed.In doingso, the dynamicsof the ossicu- auditory thresholds. Barry P. Kimberley and David A. Nelson lar chain, and to a large extentthe dynamicsof the cochlea,is removed (Departmentof Otolaryngology,2630 University Avenue SE, Room 134, from the system,leaving the eardrumcoupled to the acousticsin the ear Universityof Minnesota,Minneapolis, MN 55414) canaland in the tympaniccavity. The powerabsorption potential of the Distortionproduct emission (DPE) input-outputfunctions were eardrumfor severalinternal dissipation mechanisms isinvestigated using measuredat sevenfrequency regions, covering a rangefrom 700-6000 Hz, three-dimensionalrepresentations of the eardrum,ear-canal acoustics, for eachof 50 ears.Auditory thresholdsat eachfrequency region were andmiddle-ear acoustics. Results imply that the acousticbehavior of the determinedusing adaptive, forced-choice psychophysical techniques. middle-earair chamber(s)has a profoundinfluence on energytransmis- One-halfof the subjectshad auditorythresholds less than 25 dB HL at sionand absorptionpotential of the eardrum. everyfrequency. One-half of thesubjects had a varietyof patternsof pure sensorineuralloss. The stimuluslevel requiredto just raisean emission abovethe noisefloor is a featurethat canbe measured directly from a DPE input-outputfunction. This feature is called "emission threshold." When frequency-specificauditory thresholds were compared with correspond- 8:30 ing emissionthresholds, a closecorrelation was found over a rangeof auditory thresholdsfrom -6 to -F 55 dB SPL. A linear regression R3. Impedance measurementsin the ear canal. Sunil Puriaa• betweenauditory and emissionthreshold has a slopeof 1.0 and a regres- (Departmentof ElectricalEngineering, The City Collegeof CUNY, New sion coefficient of 0.86. However because emission threshold does not York, NY 10036) and Jont B. Allen (AT&T Bell Laboratories, 600 growas quicklyas auditory threshold at high thresholdvalues, a better Mountain Avenue,Murray Hill, NJ 07974) correlationof 0.9 is obtainedby allowinga cubicnonlinear fit. Emission thresholdscorrelated with auditorythresholds in a similarway at each The objectiveof the presentwork is to makeaccurate impedance esti- frequencyregion. Thus DPE measurementsappear to be well correlated matesat the eardrumZ,• (ca). This is doneby measuringthe thevenin with frequency-specificauditory thresholds. This findingsuggests an en- parametersof an earphoneusing a four-cavitycalibration method [J. B. hancedclinical value of DPE measurement.DPE measuresappear able to Allen, "Measurementof eardrum acousticimpedance," in Peripheral predictfrequency-specific sensory thresholds with reasonableaccuracy AuditoryMechanisms, edited by J. B. Allen etal. (1985) ] for frequencies overa moderaterange sensory thresholds. [ Work supportedby American up to 15 kHz. The calibratedearphone is then usedto measurethe ear OtologicSociety. ] canal impedanceZ• (ca) near the ear canal entrance.To transform Ze to Z,m, the ear canalcavity hasbeen modeled. The followingeffects were accountedfor by the model: (a) the jump in area from the earphone calibrationcavity to that of theear canal, (b) thedistance to theeardrum, (c) variations in cross-sectionalear canal area, and (d) effectsdue to ear canal wall impedances.It is desirablethat parametersof the ear canalbe 9:15 derivablefrom Z•. Preliminaryresults isolating each of theseeffects first in tubesand then in realear canals are presented. a• Presently,consultant R6. Spontaneousotoacoustic emissions: Evidence for a cricadianrhythm to AT&T Bell LaboratoriesAcoustics Research Department. of frequencyvariation. Holly S. Haggerty(Division of Otolaryngology, R-135, StanfordUniversity School of Medicine,Stanford, CA 94305) An investigationof the hour-by-hourfrequency stability of spontane- ousotocacoustic emissions was conducted over a 24-h period.Two emis- sionswere recordedfrom each of two audiometricallynormal adults. 8:45 Time seriesanalysis was usedto model the autocorrelationstructure of the measurements,to resolveeach 24-h seriesof measurementsinto cycli- R4. Influence of insertion depth of an insert earphone on hearing cal componentsof variousperiodicities, and to testthe statisticalsignifi- thresholdlevels and on real ear attenuation values. Frcdrik Lindgrenand canceof givencycles within thespectrum of eachseries. Results showed a Elliott H. Berger(E-A-R Division,Cabot Corporation, P. O. Box68898, significant24-h cycle for all emissions,suggesting a circadian rhythm of Indianapolis,IN 46268-0898) frequencyfluctuation. [Work supportedby the KnowlesResearch Foun- The influenceon pure-tonehearing threshold levels (HTLs) of the dation.] insertiondepth of the couplingfoam cartip of EARTONE TM3A insert audiometricearphones was investigatedin 16 subjectsusing three dis- tinctly differentinsertion depths: shallow, intermediate, and full. HTLs werealso measured using conventional earphones, model TDH-50P with MX/41-AR cushions. In addition, sound field real car at threshold (REAT) values re: ANSI S12.6-1984 were recorded in each condition. 9:30 The resultsshowed that therewas a tendencytoward poorer HTLs asthe eartip waspositioned further away from the eardrum,with the difference R7. Some implications of nonlinear passive models of spontaneous betweenthe three widely disparate insertion depths amounting to 1.3-5.3 otoacousticemissions. C. Talmadge, A. Tubis (Physics Department, dB in the frequencyrange 0.5-6 kHz. In contrast,at 8 kHz, significantly Purdue University, West Lafayette, IN 47907), and G. R. Long S44 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $44 (Departmentof Audiologyand Speech Sciences, Purdue University, 9:45 West Lafayette,IN 47907) RS. Parameterestimation for a resonanttectorial membrane cochlear model.J. B. Allen (AcousticsResearch Department, AT&T Bell Laboratories,600 Mountain Avenue, Murray Hill, NJ07974) Thenonlinear passive cochlear model of M. Furstand M. Lapid[J. Acoust.Soc. Am. 84, 222-229 (1988) ] hasrecently been criticized by Wit Oneof the basic problems ofeochlear modeling isin systematically [J.Acoust. Soc. Am. 85,2217 (1989) ] ongrounds that this model is not determiningtheparameters ofthe model from neural data. Frequency ableto accountfor theobserved statistical [ Bialek and Wit, Phys.Lett. A domainmodels give the response along the basilar membrane forpure- 104,173-178 ( 1984)]and external-tone entrainment [Long etal., Basic toneinputs, whereas the data that are desirable to matchare families of Issuesof Hearing, edited by H. Duifhuiset el. (Academic,London, 1988 ), tuningcurves as a functionof frequencyformany positions along the pp.93-100] properties of spontaneousotoacoustic emissions. The argu- basilarmembrane. This computational problem has been sidestepped by mentof Wit and the responseto it by M. Furst [J. Acoust.Soc. Am. 85, transformingtheneural tuning curve data to place, forming neural excita- 2218-2220 (1989) ] are critically examined.On the basisof numerical tionpatterns. This allows for a majorsimplification of the data fitting simulationsand exact solutions [Caughey and Payne, Int. J. Non-Linear process.If the neural data are accurately fitin the place domain for several Mech.2, 125-151(1967) ] of theFokker-Planck equation for stochasti- well-separatedfrequencies, then the frequency tuning data should fit as callydriven nonlinear passive oscillators, it is concluded,in agreement well.Using this approach, fitsto neural tuning curves using passive coch- withWit, that nonlinear passive cochlear models of the type used by Furst learmodels were greatly improved over the entire range of CFs.The and Lapid are not ableto adequatelydescribe the observedbehavior of modelassumptions willbe discussed and the results will be compared to spontaneousotoaeoustic emissions. [Work supportedby NIDCD.] measured data. 10:00-10:15 Break 10:15 with resultsof similartesting in 5- to 7-year-oldchildren [J. M. Lord- Maesand J. L. Lauter,J. Acoust.Soc. Am. Suppl.1 85, S38(1989) ] and R9. The effectof inducedround window fistula on the $P/AP amplitude adultssuggest that ABR variabilitymay be sensitiveto auditory-system ratio. Kathleen C. M. Campbell (Department of Surgery,Southern developmentalchanges that continue long after ABR absolutepeak laten- Illinois UniversitySchool of Medicine, Box 19230, Springfield,IL cieshave achieved adult values. [Work supportedin partby AFOSR.] 62794), Michael M. Savage,and Lee A. Harker (Department of Otolaryngology,University of Iowa, Iowa City, IA 52242) The ratio of the summeringpotential (SP) and actionpotential (AP) amplitudeswas studied in 19 guineapigs before and after surgicalinduc- 10:45 tion of a round window fistula. Stimuli included 2000- and 8000-Hz tone burstsand click stimuliat intensitylevels of 100,90, and 80 dB SPL. The Rll. Forward masking of the frequency following response. A. followingpvalues are for the 90-dBSPL level,but resultswere similar for K. Krishnan (Department of Communications, University of all threeintensity levels. After fistulainduction, the SP/AP ratioenlarged Pittsburgh,117 Cathedralof Learning,Pittsburgh, PA 15260) and J. in responseto the 8000-Hz tone-burst(p<0.0076) and click stimuli D. Durrant (Departmentof Otolaryngology,University of Pittsburgh, (p < 0.0001), but not for the 2000-Hz tone-burststimuli (p < 0.5627). 117Cathedral of Learning,Pittsburgh, PA 15260) Analyzedseparately, the SP significantlyenlarged after fistula induction for clicks(p < 0.0001) and 8000-Hz tonebursts (p <0.0006) but not for The frequencyfollowing response (FFR) reflectssustained neural the 2000-Hz tonebursts (p <0.5110). The AP enlargedslightly after fis- activitywithin the brain stem, phase locked to thestimulus at lowfrequen- tula inductionin responseto theclicks (p < 0.0052) but not in responseto cies.Attempts at resolvingthe issue of theorigin of FFR alongthe basilar the 8000-Hz(p < 0.1780)or the2000-Hz (p < 0.1723) tonebursts. These membranevia the subtractivehigh-pass masking technique have been resultssuggest that the SP/AP amplituderatio may be sensitiveto the unsuccessful.A tone-on-tone forward masking paradigm is describedas presenceof perilymphaticfistula with selectedstimuli. an effectivealternative method for assessingthe placespecificity of FFR. Recordingswere made between forehead and the ipsilateral mastoid using normal adult subjects.FFRs were elicitedby a 500-Hz tone burst with maskingfrequencies from 5004(0K) Hz presentedat 10 ms beforethe probe.Results indicate that the FFR (500 Hz) wasmost affected by the 1000-Hz masker,with lessmasking at 500 Hz and little or no effectsat 10:30 frequenciesabove 1000 Hz. Additionally,some subjects exhibited an en- hancementof the FFR for maskersabout 1000Hz. Resultssuggest that R10.Comparisons of between-and within-subject variability in repeated- our methodis effectivein evaluatingfrequency specificity of the FFR. A measuresauditory brain-stem responses (ABRs} in 10- to 12-year-old serendipitousfinding revealed a possiblemeans for discriminating children.Robert F. Oylerand Judith L. Lauter(Department of Speech betweenwhat mightbe separateon-set and sustained components of the and HearingSciences, University of Arizona,Tucson, AZ 85721) FFR. In reportsto thisSociety, and in variouspublications [J. L. Lauter and R. L. Loomis,Scan& Audiok 15, 167-172 (1986); 17, 87-92 (1988)], resultsof repeated-measuresABR testingin youngadults have been de- scribed,indicating that the variability of peak parameters, such as latency 11:00 andamplitude, provides information that absolute values of theseparam- etersdo not:contrasts in between-versus within-subject consistency, and RI2. The representationof syntheticstop consonants in a computational model of the dorsal cochlear nucleus. Michael J. Pont and Robert by ear of stimulation.Nine 10- to 12-year-oldboys were tested in four biweeklysessions, with fiveABR waveformscollected in eachsession for I. Damper (Department of Electronicsand Computer Science, mortauralright, mortauralleft, and binauralclicks. Relative variability University of Southampton, Highfield, Southampton SO9 5NH, measuresof ABR latenciesreveal adultlike patterns based on nonadult England) values:(1) contrastin between-versus within-subject consistency; (2) A computationalmodel of part of the mammalianauditory nervous peak differences;and (3) ear differences.There are also differencesin systemis described.This consists of threescissile stages simulating affer- within- versusbetween-session consistency. Preliminary comparisons ent neuralprocesses within the cochlea,auditory nerve, and dorsal coch- S45 d. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S45 lear nucleus(DCN). The modelderives its input from a 128-channel in orderto demonstratethe feasibilityof a "humanneuroscience" para- cochlearfilter bank. The artificialneuron units employed are simpleen- digmbased On a combinationofbehavioral testing with noninvasive neur- codermodels whose responses and interconnectionsfollow closelythose oanatomicaland neurophysiologicalexaminations of the sameindividu- reportedin recentanatomical and physiologicalstudies. The activityof als.Initial projectfocus is onbrain asymmetries related to complex-sound the low-frequencyDCN projectionunits in the modelto three (labial, processing.Preliminary results on subjectsstudied to datewill bepresent- alveolar,and velar) syntheticvoice onset time (VOT) stimulusseries is ed, documenting:( 1 ) behavioralasymmetries expressed in termsof rela- shownto mimicthe behavioralresponses of humanand chinchillasub- tive ear advantages,(2) individual differencesin anatomical(MRI) jects [P. K. Kuhl and J. D. Miller, J. Acoust.Soc. Am. 63, 905-917 asymmetries,(3) patternsof physiological"resting asymmetries," and (1978) ] to thesestimuli. The resultssuggest that DCN processingmay processingasymmetries observed during auditory stimulation (qEEG, havean importantrole in the perceptionof voicingcontrasts. [Work PET ), (4) the degreeof within-subjectconsistency of asymmetrydirec- supportedby SERCand NICHD ContractN01-HD-5-2910. ] tion and magnitudeobserved with severalnoninvasive methods (MRI, EPs,qEEG, etc.), and (5) thecoincidence of auditory-systemasymmetry descriptionsbased on behavioralversus neuroanatomico-physiological measurements.[Work supportedby AFOSR.] 11:15 RI3. Cyclic temporal pattern perception and neuronal adaptation. Donald A. Robin, Paul J. Abbas, and Linda N. Hug (Department of SpeechPathology and Audiology, The Universityof Iowa,Iowa City, IA 52242) Setsof regularlyrepeating auditory stimuli elicit uniqueperceptions; listenersare able to identifyspecific temporal patterns. Some temporal 11:45 patternsare unambiguous(only one pattern can be perceived),while othersare ambiguous(numerous patterns can be detected).While the R15. Perceptionof sequencedpitch and duration changes by patientswith psychophysicalproperties of such percepts have been well studied, little is frontoparietal lesions. Karen A. Colson (Hearing and Speech knownabout the underlyingneurological bases of temporalpattern per- Department,University of KansasMedical Center, KansasCity, KS ception.In this experiment,the role of adaptationin temporalpattern 66103),Donald A. Robin,and Erich S. Luschei(Department of Speech perceptionwas examinedby studyingneural responsesin four catsto a Pathologyand Audiology, University of Iowa, Iowa City, IA 52242) temporalpattern that isperceptually unambiguous and one that is percep- Currentevidence suggests that auditoryprocessing of nonlinguistic tually ambiguous.Measurements were made of the whole-nerveaction pitchand timing features may be differentiallyimpaired following focal, potential,the auditory brain-ste mresponse, and potentials from the sur- unilateralcortical insult. Three experiments were conducted to examine face of the primary auditorycortex. The adaptationpatterns corre- the abilityof patientswith well-defined,left, frontoparietallesions and spondedwith the perceptual organization of temporalpatterns in humans normal,matched controls to perceiverelative pitch and/or duration at all levelsof the nervoussystem studied. changesin tonalsequences. Performance patterns for two right frontopar- ietalpatients also were examined. Subjects identified whether or notany tonein a three-tonesequence was higher in pitchand/or longer than the othertones. Compared to the controls,the left anteriorgroup typically 11:30 requireda significantlylarger difference (e.g., Hz and/or ms) between RI4. Functionalorganization of normal humanauditory central nervous thestandard and comparison tones to accuratelyidentify relative changes systemsobserved with multiple noninvasivetechniques: Year one of the in pitch,duration, and pitch and duration combined. The rightfrontopar- CN$ project.Judith L. Lanter (Departmentof Speechand Hearing ietai patientsperformed the three tasksat a level similar to that of the Sciences,University of Arizona, Tucson, AZ 85721) normal controls.Based on the findings,the perceptiontasks seem to in- volvea complexinteraction between temporal and spectralfeatures and The coordinatednoninvasive studies (CNS) projectis designedto appearto besensitive for detectingleft, anterior patients' difficulties with bringtogether a varietyof noninvasivemethods for studyingliving brains theseaspects of auditoryprocessing. S46 d. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S46 WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM D, 8:15 TO 11:59 A.M. SessionS. SpeechCommunication IV: Memorial Sessionfor Dennis Klatt StefanieShattuck-Hufnagel, Cochairman ResearchLaboratory of Electronics,Room 36-51L MassachusettsInstitute of Technology,Cambridge, Massachusetts 02139 Kenneth N. Stevens, Cochairman ResearchLaboratory of Electronics,Room 36-517, Massachusetts Institute of Technology,Cambridge, Massachusetts 02139 Ralph N. Ohde, Cochairman VanderbiltUniversity School of Medicine,Division of Hearingand SpeechScience, Box 562, Station17, Nashville, Tennessee 37232 Invited Papers 8:15 S1. Acousticsand synthesis of fricativeconsonants. Kenneth N. Stevens(Research Laboratory of Electronics and Departmentof ElectricalEngineering and ComputerScience, Massachusetts Institute of Technology, Cambridge,MA 02139) This paperattempts to integratesome recent research of Dennis Klatt on the analysisand synthesisof fricativeconsonants with further experimentaland theoreticalstudies of fricativeproduction. The speech productionstudies examine data on airflowsand pressuresduring voiced and voicelessfricatives, and estimate fromthese data the time variation of theareas oft heglottal and supraglottal conslriclions and the spectra of the soundsources at theseconstrictions. These calculations are basedon theoreticaland experimentaldata on airflow in constrictedtubes and on soundgeneration in turbulentflow. Acousticspectra in a numberof utterancescontaining fricatives in variousvowel environmentsare measuredat critical pointswithin the utterances,and are interpreted in termsof theproduction studies. Particular at tention is paidto eventsnear the consonant-voweland vowel-consonantboundaries, where the dominantsource changes from fricationnoise to aspirationnoise to glottal vibration.Based on this research,new synthesisrules for fricativeconsonants are proposed.[Research supported in part by NIH Grant No. NS-04332.] 8:40 S2.Dennis Klatt's contribution to automaticspeech recognition. Victor W. Zuc ( RoomNE 43-601A, Spoken LanguageSystems Group, Laboratory for Computer Science,Massachusetts Institute of Technology, Cambridge,MA 02139) Over the past20 years,Dennis Klatt hasmade enormous contributions to the fieldof automaticspeech recognitionthrough his research, writing, and student supervision. In theearly 70's, he servedas a memberof thesteering committee of theARPA SpeechUnderstanding Research (SUR) program,providing leadership and guidanceto the researchcommunity. He alsoparticipated actively in speechrecognition research, first performinga setof spectrogramreading experiments assessing the role of varioussources of knowledge,and laterinvestigating the useof synthesis-by-ruletechniques for wordverification. Out of thisinvolvement with the ARPA-SUR programcame the landmarkpaper reviewing its technicalachievement. as well as several publicationsdescribing his own proposals,LAFS and SCRIBER, for humanand machinespeech recognition. Over the past 10years, Dennis directed his attention to the designof signalrepresentation fronl-ends, as well as the investigationof perceptuallymotivated distance metrics in order to implementhis speechrecognition models.This talk paystribute to Dcnnis'sincredibly active research life by examiningthe legacyhe left behind in speechrecognition. 9:05 $3. Duration modelsand segmentalquality in a text-to-speechsystem. Roll Carlson and Bj6rn Granstr6m (Departmentof SpeechCommunication and MusicAcoustics, Royal Instituteof Technology,Box 70014, S- 100 44 Stockholm, Sweden ) We havehad the privilegeof workingtogether with DennisKlatt for manyyears. In our presentationwe will referto someof Klatt's work that hashad an influenceon our own work.Modeling of segmentalduration S47 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S47 wasa centralpart in Klatt's work duringthe 1970s.This work resultedin a durationmodel in 1979that capturesmany of thebasic effects found in speech.This modelhas been used as a frameworkin the KTH text- to-speechsystem. The useof quantityin Swedishdemands expansions of the model.The Swedishstudy has beendone in the contextof a speechdatabase from differentspeakers reading different text materials.In Klattalk, Klatt [J. Acoust.Soc. Am. 82, 737-793 (1987) ] addressedall levelsin a text-to-speechsystem, but specialeffort was placed on a generalimproved segmental quality. The qualityof the bestspeech synthesis is, however,still far from that of humanspeech. Some recent efforts to improvethe segmentalintelligibility in our systemwill be described.This includesexperiments with newsynthesis strategies with an emphasison model- ingcontextual variability. Analysis and synthesis of positionalvariants of theSwedish consonants are reported, and new strategiesfor synthesisare discussed.Based on analysisof the speechdatabase, consonant rules affectingboth sourceand resonatorfeatures were formulatedand tested.Special efforts were madeto handle the realizationof consonantclusters. In thisdevelopment work, diagnostictests were usedat regularintervals. Resultsfrom the last years'evaluation will be reportedand discussed. 9:30 S4. Perceptualevaluation of MITalk and DECtalk. David B. Pisoni (SpeechResearch Laboratory, Departmentof Psychology,Indiana University, Bloomington, IN 47405) In the springof 1979,we beganthe firstof what wouldeventually become many dozens of behavorial studieson the perceptualevaluation of syntheticspeech produced automatically by rule. The MITalk text-to- speechsystem was just nearing completion at MIT. At thattime, I spentmany hours with Dennis Klatt talking aboutand planning a varietyof perceptualtests to evaluatethe MITalk system.In thispaper, I will first summarizethe initial resultsobtained with the MITalk systemin 1979.Then I will describethe evaluationof DECtalk. I believethat one of the reasonsthat DECtalk hasconsistently shown such high levels of segmental intelligibility,levels often approachingthose observed with naturalspeech, was Dennis'intense fascination withthe results of ourerror analyses of theMRT thathe used to selectivelymodify, refine, and improve the qualityof thesynthetic speech produced by DECtalk. In workingwith Dennis, it becameclear to me that one of hismajor goals was to develop a systemtht wouldproduce the very best quality synthetic speech possible. The datafrom additional perceptual tests demonstrate clearly that Dennis was successful in achieving his goals for DECtalk.DECtalk remains the standard against which all othertext-to-speech systems are compared. [ Work supportedby NSF. ] 9:55 S5. Adultsand infantsshow a "prototypeeffect" for speechsounds. Patricia K. Kuhl (Departmentof Speech and Hearing Sciences,University of Washington,Scattic, WA 98195) DennisKlatt providedvaluable assistance to manyspeech perception researchers. His adviceon the syn- thesisof speechsignals was particularly helpful, and investigators relied on hisexpertise when problems arose in thedevelopment of stimulidesigned to approacha problemin a newWay. This was the case in theprepara- tionof stimulito testwhether there arc "prototypes" for vowelsounds. The experimentalquestion was whether or not adults and infantsresponded differently in a within-categoryvowel discriminationtask when the "standard"stimulus was an exceptionallygood instance of the vowel/i/--a prototypeof tbe category--as opposedto a nonprototypic/i/ vowel.The resultsshowed that whenthe prototype oftbe categoryserved as the standardstimulus, it was more difficult to hear differencesbetween it and novel/i/vowds than it was to hear differencesbetween the nonprototypicstimulus and novel/i/vowels. In other words, the prototypewas perceivedto be moresimilar to newmembers of the categorythan was the nonprototypicstimulus. The effect wasobserved both in adultsand in 6-month-oldinfants. Described first will be the hypothesisunderlying the prototypetest, the methodused to constructthe stimuli(the teststimuli for the prototypeand the nonproto- type were scaledpsychophysically using the "mcl" scale), and the resultson Englishadults and infants. Describednext will be the secondphase of the researchprogram, which entails cross-language tests on Swedish adultsand infants. The cross-languagetests are designedto assesstwo differentexplanations for the prototype effect:( 1) that particularvowel stimuli (/a,i,u/) are inherentlymore resistent to the effectsof articulatory/ acousticchange (Stevens' quanta] theory) and (2) thatthe effects observed in Americanadults and infants are attributableto experiencein listeningto English,even in the first6 monthsof life. Carefulpreparation of the Englishand Swedishstimuli was critical in designingthese experiments. Dennis Klatt wasalways ready to provideadvice on such matters; his assistance isgratefully acknowledged and sorely missed. [ Work supported by NIH. ] 10:20-10:35 Break ContributedPapers 10:35 One type of articulationdisorder is a sounddistortion that has been definedas an allophonicvariation within the perceptualboundary of a S6. Perceptual and acoustic charactertics of distorted /r/. Ralph target phoneme.An establishedfinding in speechperception is that N. Ohde, Michael E. McCarver (Division of Hearing and Speech soundsare more accurately identified across sound categories than within Sciences,Vanderbilt University School of Medicine, Nashville, TN soundcategories. In order to determineif distorted/r/could be accurate- 37232), and Donald J. Sharf (Boca Raton, FL 33496) ly and reliably perceived,six speechpathologists identified the produe- $48 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S48 tionsof prevocalic /r/ and/w/ wordsof 12children diagnosed ashaving timingof supraglottaland glottal movements of/h/. Utterancesin which an/r/misarticulation. The results of the identification tests revealed a /h/was presentor absent(e.g., "new heart" versus"new art") were relativelyhigh average distorted/r/category of 70% or betterfor four analyzedacoustically and contrasted.The corpusconsisted of about20 children.Moreover, intrasubject reliability scores for thesedistorted/r/ suchutterances repeated several times by threespeakers. The acoustic childrenaveraged 80% or better.Preliminary findings of spectrographic datashowed evidence ofbreathy voicing at the/h/-vowelboundary in all analysesof formant transitiononsets show that F3 onsetsof distorted/r/ cases,and that generationof turbulencenoise during the consonant oc- aresubstantially higher than F3 onsetsof/r/for normaland synthetic curredboth in thevicinity of theglottis (aspiration noise) and the vicinity versionsof children'sspeech. [Work supportedby BiomedicalResearch of the supraglottalconstriction (frication noise). The relativecontribu- SupportGrant No. RR-05424.] tionof thetwo noise sources depended on the vowel, with greaterfrication noiseoccurring for high vowels. When an/h/was in positionbetween two vowelsor glides,it generallyadded little or no durationto the utterance, relativeto the contrastingutterance with no/h/. Implicationsfor the 10:47 phonologicalstatus of/h/as discussed.[ Work supported by NIH grants to MIT. ] S7. Stress shift as the placementof phrase-levelpitch markers. StefanieShattuck-Hufnagel (36-511 MassachusettsInstitute of TechnologyResearch Laboratory of Electronics,77 Massachusetts Avenue,Cambridge, MA 02139) TheF0 contoursproduced by the text-to-speech conversion program 11:23 Klatttalk[D. H. Klatt, J. Acoust.Soc. Am. 82, 737-793(1987)] are basedon a translationIS. Maeda,RLE-QPR 114,193-211 (1974) ] for S10, Percel•tion of some consonantcontrasts in noise. Abeer Alwan AmericanEnglish of the "hat pattern" approach developed for Dutch [J. (ResearchLaboratory of Electronicsand the Departmentof Electrical 't Hart andA. Cohen,J. Phon.1, 309-327(1973)]; thisapproach is Engineeringand ComputerScience, Room 36-511, Massachusetts similarto anearlier description by Mattingly[I. G. Mattingly,Supple- Instituteof Technology,Cambridge, MA 02139) mentto HaskinsLaboratory Status Report on SpeechResearch, 1-223 Thegoal of the present study is to examine the acoustic properties that (1968)]. Onequestion that arises for this view ofF0 patternsishow might listenersuse to dislinguishbetween speech sounds when these sounds are it dealwith the phenomenon of "stress shift": For somespeakers, the presentedin noise.A seriesof perceptualexperiments was conducted prosodicprominence on the main-stresssyllable of wordslike "thirteen" usingnatural stimuli consisting of nonsenseCV syllables,where C was and"Mississippi" isperceived to moveto an earliersyllable when these either/m/,/n/,/b/, or/d/, and V was either/o/,/i y/, or/o '•/. The wordsappear in phraseslike "thirteen men" and "Mississippi mud." This stimuli were degradedby addingvarious levels of white noiseand were paperwill report pitch and duration measurements designed to evaluate presentedto subjectsin identificationtests. Preliminary results show that thehypothesis that at leastsome aspects of thestress shift phenomenon when the noise is at a level such that the transition of the second formant canbe described as the simple placement of theonset rise of a hatpattern frequencyof thevowel is masked, confusions between the place of articu- on an earlysyllable of the prosodicphrase. lation for the stimuli occur. Noise levels for which confusions in manner of articulationoccur can also be predictedfrom maskingtheory. These resultsare compared with resultsreported earlier [e.g., Miller andNicely (1954) ] where the thresholdsof idenfication were describedin terms of 10:59 thesignal-to-noise ratio. These results will bediscussed further in termsof the acoustictheory of speechproduction and the maskingtheory of the S8. Synthetic speech audiomerry, Cotinc Bickley (Sensimetrics auditorysystem. [Work supportedin part by an NIH grant.] Corporation,Building 100, One Kendall Square,Cambridge, MA 02139) and Gerald Kidd (Department of CommunicationDisorders, Boston University, Boston,MA 02215) A newhearing test is beingdeveloped that is basedon presentingto listenerssets of synthesizedwords with well-definedacoustics properties. The testis based in parton workby G6syet al. [ Proc.11 th ICPS, Tallinn 11:35 (1987) ], and its aim is to estimatea listener'shearing sensitivity from errorsin worddiscrimination. Sets of wordshave been synthesized ( using Sll. The influenceof selectedacoustic cues on the perceptionof/1/and theKlatt synthesizer)that differ from each other by one of twophonemes /w/. Carol Y. Espy-Wilson (Research Laboratory of Electronics, (e.g.,sit sat fat fit ). The synthesiswas guided by two goals.( 1) Eachword MassachusettsInstitute of Technology,Cambridge, MA 02139) shoulddiffer from anotherword in a setby only oneacoustic feature; the primarydifference must be limitedto a specificfrequency band (e.g., sit In a semivowelrecognition system developed by Espy-Wilson[Mass. versussat differ by the frequencyof the firstformant). (2) The synthe- Inst. Technol.Res. Lab. Electron.Tech. Rep. No. 531 (1987)], the sizedwords should be highly intelligible to normallyhearing listeners in a sounds/l/and/w/were frequentlyconfused, especially when they oc- quietenvironment. Initial resultswere obtained by presentingthe synthe- curredintervocalically. In thisstudy, the perceptualimportance of some sized words combined with white noise of various levels in a forced-choice of the cuesused in the recognitionsystem, as well as someothers which paradigmto normally hearing subjects,.As expected,word discrimin- appearto be salient,were investigated.An [ala]-[awa] continuumwas ability wascorrelated with the salienceof the acousticfeature that distin- synthesized.The startingpoint wasan easilyidentifiable [ala] stimulus. guishesthe word relative to theadded noise. The feasibilityof usingsyn- Three factorswere variedorthogonally to shift the percepttowards thesizedword setsof this type to detectand estimatethe severityof [awa]: (1) the rate of changein the formanttransitions between the hearingimpairment will bediscussed. [Work supportedby a grantfrom semivoweland followingvowel, (2) the rate of changein the amplitudes NIH.] off 3,F 4, andF 5 betweenthe semivowel and following vowel, and ( 3 ) the spectralshape of the semivowel(coronal or labial). Preliminaryresults of anidentification test show that spectral shape and the rate of changeof the formanttransitions are important cues, whereas the rate of changein the amplitudesofF3, F4, andF5 appearsto havea negligibleeffect on listen- 11:11 ers' responses.For example,with formant transitionsof 30 ms or less, S9. Acoustic properties of /h/. Sharon Y. Manuel and Kenneth [ala] is heard.With formanttransitions greater than 40 ms,the percep- N. Stevens (Research Laboratory of Electronics, Massachusetts tion moves towards [ awa]. The results also show that a few listeners had Instituteof Technology,Cambridge, MA 02139) difficultyhearing [awa] whenthe semivowel had a coronalshape, despite the rate of changein the formanttransitions being biased towards/w/. Theaim of this paper is twofold: ( 1) to investigatethe physical mecha- Theseresults will bediscussed with respect to pastand future research in nismsof soundgeneration for the consonant/h/and (2) to examinethe speechrecognition. S49 d.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S49 11:47 addressedhere: Development of newsystems for controlofF0 and dura- tion,and mechanisms for addingvariety. The proposedsynthesis system S12.Combining statistical and linguistic models for synthesisof prosodic is a statisticalmodel trained on text, partsof speech,pronunciation, lexi- tontours. Marl Ostendorf (Boston University. ECS Department, Boston.MA 02215). Patti Price (SRI International, Menlo Park, CA cal stress,prosodic labels (major and minor boundaries,accents, etc.), and acousticparameters (relative F0 and duration). The synthesisprob- 94025), StefanieShattuck-Hufnagel (Massachusetts Institute of lemis to predictthe prosodic labels and acoustic parameters given the text Technology, Cambridge, MA 02139), NanetteVeilleux, and the statisticalmodel. Several hours of speechhave been collected from Colin Wightman.and Rudy Garcia (BostonUniversity. Boston. MA 02215) professionalFM newscasters,a labeling scheme has been converged on, anda portionof the data hasbeen labeled. The componentsof the system "It is very importantto get the timing,intonation, and allophonic sofar implementedwill bediscussed: ( 1 ) statisticalmodeling of sequences detailcorrect in orderthat a sentencesound intelligible and moderately of partsof speechto predictmajor prosodic breaks, (2) the roleof breath natural." [D. Klatt, J. Acoust.Soc. Am. 82, 737-793 ( 1987}l. This im- noisein naturalhesS,and (3) the implementationof a sinusoidalmodel for portantreview article included prosody as a researchissue for improving duration and pitch modificationof waveforms.[Work supportedby text-to-speechsynthesis. Klan's suggestionsfor improvingprosody are NSF.] WEDNESDAY MORNING, 29 NOVEMBER 1989 DIRECTORS' ROW 43, 8:30 A.M. TO 12:00 NOON SessionT. Structural Acousticsand Vibration II: Characterization of Viscoelastic Polymers Wayne T. Reader, Chairman VectorResearch Company, Inc., 6903 RockledgeDrive, Suite 1200, Bethesda,Maryland 20817 Chairman's lntroduetion•g:30 Invited Papers 8:35 TI. Measurementof polymercomplex modulus properties using several techniques. Thomas M. Lewis (Anatrol Corporation,10895 Indcco Drive, Cincinnati,OH 45241) and DominiqueLcgros (Metrarib Instruments, France) Varioustechniques for measurementof the dynamicmechanical properties of polymersare discussed. Specialemphasis is givento thevibrating beam, SDOF resonance, SDOF impedance, creep, relaxation, forced oscillation(Metrarib viscoaualyzcr),and forcedtorsional oscillation (Metrarib micromechanalyzer)tech- niques.Material propertydata, generated over wide ranges of temperatureand frequency(using the above techniques),are presentedin termsof reducedfrequency upon application of an Arrheniustemperature- frequencyshift function.Comparison of theseproperties in the reducedfrequency format indicates good correlationbetween techniques. Specific examples are included consisting of datagenerated from commercial- ly availablematerials in bothshear and tension-compression states of stressfrom which e•timates of Poisson's ratio may be determined. 9:00 T2. The ASTM E-756 DampingStandard---The good, the dangerous,and the pitfalls.Michael L. Drake (Universityof Dayton,300 CollegePark JPC-36,Dayton, OH 45469) The resonantcantilever beam test procedure is a reliabletest method from whichcomplex modulus data canbe determined. The ASTM E-756-83Standard was developed around this test procedure. This testmethod hasbeen used since Oberst started work in the early 1950s.Ross, Kerwin, andUngar expanded the usefulness of the testthrough the developmentof sandwichequations. Although the testprocedure is reliable,it was developedwhen the fundamental interest in complexmodulus data was focused on the transitionregion. As a result,when users begin to stretchthe viseoelastic properties testing further into the rubberyand glassy regions problemsdeveloped. This paperwill detailthe idiosyncrasiesof this test method and the analyticalequations usedto calculatethe complexmodulus data from the raw testdata. It will be demonstratedthat apparently good,self-consistent data can be very inaccurate. The effectsof testspecimen configuration and the modulus of the viscoelasticmaterial on the accuracyof the complexmodulus data and the utility of the varioustest specimenswill bediscussed. The cornmortality of theproblems in thebest test with other popular test methods will alsobe discussed and suggested test parameters will begiven to enhancecomplex modulus data accuracy. S50 J. Acoust.Soc. Am. Suppl. 1, VoL86. Fall1989 118thMeeting: Acoustical Society of America S50 9:25 T3. Frequency-temperaturesuperposition in polymer damping behavior. David I. G. Jones(Materials Laboratory,WRDC/MLLN, Wright-PattersonAFB, OH 45433) Theproblem of determiningoptimum shift factors for frequency-temperature dependence of polymeric materialcomplex modulus properties has not been fully resolved even halfa century after the pioneering work of Williams,Landel, and Ferry. The question of whatform, such as the WLF (Williams,Landel, Ferry) equationor theArrhcnius equation, best depends on both the material and th& quality of the test data. In most cases,the data scatter is sufficient to prevent a definitivechoice. These issues will bediscussed with reference to availabletest data for severalpolymcric materials, obtained by various measurement techniques. It will be shownthat the least ambiguous shift factor estimates are obtained when data scatter is as low as possible, and thefrequency range of dataat eachtemperature isas wide as possible, as would be expected. It will also be shownthat differences arising from the use of differentshift factor equations arc not sufficient for most engineeringapplications. 9:50 T4.Presentation andmodeling ofcomplex modulus. Lynn Rogers (WRDC/FIBG, Area B, Building 24C, Room220, Wright-PattersonAFB, OH 45433-6503) Oneof thehistorical challenges in the field of passive vibration damping technology isthe modeling of experimentalcomplex modulus data as a functionof temperature and frequency. Several models exist for the temperatureshift function (TSF); i.e., Arrhenius, WLF, exponential, etc. Several models of complex modulus asa function ofreduced frequency also exist. All existingTSF and CM models fail to represent atleast some sets of datawith desired accuracy and efficiency. Consider a linear, constant coefficient, stable system, and its frequencyresponse function. It is well known that if thereal component ofthe complex-valued frequency responsefunction is given over the infinite frequency range, then the imaginary component may be obtained. Thecomplex modulus of vibration damping materials issuch a system.Extensive work with fractional calculus basedmodels for complexmodulus has established their viability and potential attractiveness. A ratio of factoredpolynomials ofone-half order is proposed tomodel the complex modulus. This CM modelis attractive froma numberof viewpoints: The proper interrelationship ofthe real and imaginary components isguaran- teed;an adequately large number of terms may be used in order to accurately model the complex modulus; an expressionmay be developed forthe real component that lends itself to fitting data by collocating through a numberof points; closed-form expressions may be developed forcompliance, relaxation modulus, and creep compliancewhich also lend themselves tocollection fitting of experimental data, etc. With modern computa- tionalpower, this model becomes both accurate and efficient. Previous work has established theslope of the TSFas the characteristic which causes complex modulus data to be properly shifted; therefore, anew approach tomodeling the TSF is proposed. The new model is based on determining values of slopes at equallyspaced temperatures,fitting a cubicspline through these points (i.e., knots), storing the coefficients, and integrating thecubic spline analytically. The concept of reducedtemperature isintroduced, used as a conveniencefor the presenteffort and proposed as an additionalmethod of presentingdata in a formuseful to thedamping industry.The core of the revolutionary concept isusing the simultaneous modeling ofboth real and imaginary componentsasthe criteria to enablethe set of datato establishits TSF. Previous techniques have used real modulus,imaginary modulus, and loss factor as a functionor reduced frequency, sometimes ina least-squares sense,and sometimesvisually, as the criteria.The aboveCM and TSF modelsare essentialto the iteration strategyrequired to determine parameter values for both models. The iteration scheme isconceptually straight- forward.Approximations tothe CM and to the TSF are obtained. For each TSF knot, the reduced temperature isused to determine the associated reduced frequency, the current loss factor curve is compared tothe corre- spondingexperimental value and the value of the slope adjusted accordingly, the real component collocated for theupdated TSF, etc. Examples are given and discussed. 10:15 TS. Designof polymersfor viscoelasticdamping applications. R. E. Wetton(Polymers Laboratories, Ltd., The TechnologyCenter Epinal Way, LoughboroughLEll 0QE, UnitedKingdom) and J. L. Duncan (PolymersLaboratories, Inc., AmherstFields Research Park, 160Old Farm Road,Amherst, MA 01002) The largefrequency dependence of the dynamic moduli of polymersallows the design of materialswith specialdamping and engineering applications. The temperature dependence of these properties is, however, a drawbackin manycases. The presentpaper reviews the changesof Youngs,shear, and bulk moduliwith frequencyand temperature. Their accurate measurement and predictions via time/temperature superposition methodsare discussed, as are therelationships between these parameters with changingtemperature. The designofelastomers with high damping and minimized temperature variation is discussed with some examples. 10:40 T6. Internal friction in polymer systems.Jozef Bicerano and James K. Rieke (Materials Scienceand DevelopmentLaboratory, Central Research, The Dow ChemicalCompany, Midland, MI 48674) Internalfriction is an effective mechanism for dissipating energy in polymersystems. Of particularinterest is the ability to dampenand absorbunwanted acoustic and mechanicalvibrations. The useof internalfriction S51 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America 551 will becomemore effective when detailed geometric, thermodynamic, and kinetic models are developed for the physicalphenomena producing the vibrationdamping. A modelthat relatesthe viscoelasticproperties of polymericsystems to theirmolecular level structures has been developed. This model considers the nature of thetemperature an• frequencydependences of the storage and loss components of the complex moduli of polymers.It providescorrect and internally consistent correlations. Therefore, it constitutesa first step in an attemptto bridgethe gap between the results of applicationof an externalstress (such as acoustic or mechani- cal vibration)and the molecularlevel properties of the polymers. 11:05 T7. Characterizingviscoelastic materials using the free volumemicroprobe. Bret A. Mayo, JamesP. Pfau, and DuryodhanMangaraj (Battelle Memorial Institute, 505 King Avenue,Columbus, OH 43201-1693) A greatdeal of effort has been directed toward the development of newpolymer blends and interpenetrating polymernetworks in recentyears. This approachhas emerged as an importantmeans of developingnew polymcricmaterials and improving properties such as sound damping. The thermodynamicmodels that best describepolymer-polymer interactions have an importantfree volume contribution. Until recently,however, therewas no techniquethat couldmake a direct,nondestructive measurement of the molecularfree volume. Usingthe free volume microprobe (FVM), it isnow possible to characterizeboth the average free volume site sizeand the relativenumber of freevolume sites. Thus it is possibleto makesome assessment of the total free volumeas well as the freevolume distribution. The basictheory of the FVM techniquewill bcpresented as well assome supporting data from a seriesof miscible,immiscible, and partially misciblc polymer blends. 11:30-12:00 OpenDiscussion WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM H, 8:30 TO 11:35 A.M. SessionU. Underwater AcousticsIII: Computational AcousticsI Henrik Schmidt, Chairman Departmentof OceanEngineering, Massachusetts Institute of Technology,Cambridge, Massachusetts 02139 Chairman's Introductions8:30 Invited Papers 8:35 U1. Inversionusing simulated annealing. Atanu Basu and L. Neil Frazer (HIG, University of Hawaii Manoa, Honolulu, HI 96822) Theproblem here is finding the profile • thatminimizes E(•) = II• - •(•)II - •ll• - •ollin which • is a data set, •0 is the bias, and • is a Grccn's function. If• is nonlinear and the dimensionof • is large, then findingE,,•, is difficult.Simulated annealing (SA) seeksE•.•, by samplingfrom the probabilitydistribution p(r•) = exp[ - E(•)/T]/Z in which T is so small that p spikesat E,,•,. This wouldbe silly if one had to generateZ numerically,but SA doesnot needto knowZ. The physicalanalogy for SA is that eachcomponent of• isan atomin a melthaving temperature Tand freeenergy E. Rapidcooling of the meltgives a glass(high E) but slow coolinggives a singlecrystal (low E). Problemswith SA are that slow coolingconsumes the computerbudget faster than it findsE•.i, and that the freezingtemperature T½ is difficultto determine. Accordingly,much rescatch on SA concernscooling schedules that do not rcquireT c. In wavcfield inversion problems,it wasfound that knowncooling schedules gave poor results and that knowledgeof T½was essential. A rapid methodto find Tc wasdiscovered. Also, SA wasno longerthought of as a processthat terminates; rather T½is determinedaccurately keeping the melt at T½and lettingSA find manygood •. A measureof resolutionin• spaceisthe "order parameter" •/= II• - •ll, where•t isthe unknown true profile. Expected valuesof v(E) canbe obtained from syntheticdata. S52 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S52 8:55 U2.Adaptive array processing of full-wave acoustic fields. Michael S. Wcngrovitz(Atlantic Aerospace ElectronicsCorporation, Waltham, MA 02154) and Hcnrik Schmidt(Massachusetts Institute of Technology,Cambridge, MA 02139) Adaptivearray processing methods have been developed andapplied ina numberofareas including radar, communications,and underwater acoustics. Assessing the performance of thesemethods is moredifficult in theunderwater acoustics context due to the complex influence ofthe ocean environment onthe signal and noise fields.Arrivals across the array are not plane waves, as typically assumed in radarand communications problems,but are solutions tothe wave equation applicable tothe particular ocean environment. This paper describesanefficient computational method that interfaces the full-wave solution for horizontally stratified mediawith adaptive array processing algorithms. The computational method has been used to provide coher- entarrival data for full-wave array beam pattern calculation, and noise spatial-correlation matrices for deriving arraygain corresponding to various adaptive arrays. The computational method has been used to studya numberof importantarray processing issues for multidimensionalarrays with large number of elementsin oceanicenvironments. Example issues discussed in the paper include applicability ofpartially adaptive beam- formingalgorithms, sensitivity ofarray performance toenvironmental parameters such as bottom composi- tion,convergence rates of adaptive nulling algorithms in underwaterenvironments, and array calibration. 9:15 U3.Recent developments in parabolic equation modeling. Michael D. Collins( Naval Research Laboratory, Washington,DC 20375) Theparabolic equation (PE) is oneof themost important computational models for underwatersound propagation.Since the PE was introduced into the field of underwater acoustics by F. D. Tappert,a greatdeal ofresearch has been devoted to extending itscapabilities, accuracy, and efficiency. In its present state, the PE is capableof handlingthree-dimensional problems, rough interfaces, and interaction with elastic ocean bottoms. A higher-orderPE is accuratefor arbitrarilylarge propagation angles. The rotatedPE, whichmarches the solutionparallel to theocean bottom, provides improved accuracy for problemsinvolving sloping bottoms. Theefficiency ofthe finite-difference solution of thePE has been improved for range-dependent problems with a newalgorithm for solvingtridiagonal systems. The time-domainPE efficientlysolves pulse propagation problemsdirectly in thetime domain. Some recent progress aswell as some unresolved issues in PEmodeling will be discussed. ContributedPapers 9:35 transforms(FFTs). Preventingsignificant errors due to theperiodicity of the finiteFFT ("ringing") requiresinserting an absorberpotential to u4. Parabolicequation starting field for a low-frequencysource near an confinethe wave energyto the rangeof interest.Such potentials, or interface.Kenneth E. Gilbertand Dehua Huang(National Center for sponges,have, to date,been strictly ad hoc.The problemof creatingan PhysicalAcoustics, University, MS 38677) optimalsponge by usingan opticalmodel analogy of the reflectionless Experimentsin underwateracoustics often usea low-frequency potentialU = U,* [sech(a.x)]•-of 1-D quantummechanics is studied. sourceclose to the bottom. In such situations,the standardGaussian This approachpermits the developmentof closed-formexpressions for startingfield for parabolicequation calculations can grossly violate the bothtransmission and reflection coefficients in termsof elementarytran- boundaryconditions at thewater-bottom interface. In thispaper, a sim- scendentaland gamma functions of complexargument. Using these coef- pleoperator method is derived that gives the correct boundary conditions ficientsand an efficienttechnique for evaluatingsuch gamma functions in thestarting field a0d also properly excites the "high-angle" modes and yieldsa quantitativestudy of the problemof theoptimal sponge. These thecontinuum. As limiting cases, the standard Gaussian starting field of resultsdemonstrate that, for relevantparameter domains, there may be Brock[ NORDA Tech.note 12, Naval Ocean Research and Development several"branches" of minimalpaths in spongeparameter space, includ- Activity,NSTL Station, MS (1978)] andthe high-angle starting field of ingone particularly well suited to thedemands of PE modeling.[Work Greene[I. Acoust.Soc. Am. 76, 1764-1773(1984) ] areobtained analyti- supportedby NOSC-IR and OCNR.] cally.The methodused here is comparedwith numericalresults from a fast-fieldprogram for a low-frequencysource near the bottomin shallow water. 10:05 U6. Modal decomposit|onof fieldsin range-varyingwaveguides. David J. Thomson(Defence Research Establishment Pacific, FMO Victoria, 9:50 British Columbia V0S lB0, Canada) U5. Optimal absorberpotentials in PE modeling.F. J. Ryan and C. Numericalsolutions to parabolicequations (PE) providethe total David Recs(Code 541, Naval OceanSystems Center, San Diego, CA contributionto the field at each point on the computationalgrid. For 92152-5000) someapplications, it is instructive to examinethe propagation of individ- ual spectralcomponents, particularly in range-dependentwaveguides, The parabolicequation (PE) approximation---essentialin solving wheremode coupling can be important. This paper presents a novel meth- range-dependentpropagation problems--relies on iteratedfast Fourier od for decomposinga numerical PE fieldinto its horizontalwavenumber $53 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1089 118thMeeting: Acoustical Society of America 553 spectrum.At a givenrange in a range-dependentwaveguide, the PE field downslopepropagation show that the modifiedwave equation performs versusdepth is providedas initial data to a PE modelwhich is usedto verywell with errors being reduced by 75% forrealistic slopes and density propagatethe field in a range-independentwaveguide characterized by the contrasts. local environmentalconditions at the rangeof the initial data. A Fourier transformof this propagatedfield yieldsthe PE modalspectrum directly. Moreover,if the standardparabolic equation is usedto propagatethe field,then the PE modalspectrum can be postprocessedinto the correct modalspectrum for the one-waywave equation [D. J. Thomsonand D. 10:50 H. Wood, J. Acoust. Soc. Am. 82, 224-232 (1987) ]. A numericalexam- U9. A convolutionalgorithm for standardparabolic-equation modeling. ple involvingupslope propagation is usedto illustratethe proposedspec- Rober•F. Joyceand Martha E.M. Head (Naval OceanographicOffice, tral decompositionmethod. StennisSpace Center, MS 39522-5001) The split-stepFourier algorithmfor solvingthe acousticparabolic equationinvolves two Fouriertransforms, one forward and the otherin- 10:20 verse.These may, of course,be replacedby a convolution.For the stan- dardparabolic equation, the particularform of oneof the convolvedfac- U7. Boundary integral method for acoustic propagation in laterally torsallows the convolutionintegral to be solvedusing a singleFourier inhomogenous medium. Woojoe Seong and Henrik Schmidt transform.Application of this techniqueresults in a limitationon the (MassachusettsInstitute of Technology,Cambridge, MA 02139) allowedrange step. When applicable, however, the convolution algorithm reducescomputer run timesto about60% of thoserequired by the stan- A hybrid analytical-numericalmethod that combinesboundary ele- dard split-stepFourier algorithm. Examples are presentedand limitations ment and full-wavesolution technique is presentedfor two-dimensional are discussed. acousticpropagation in laterallyinhomogenous and simultaneously verti- cally stratifiedmedium. First, the range-dependentocean is dividedinto range-independentsectors and the field within eachsector is expressedby a boundaryintegral over the verticalsector boundaries in termsof a setof unknownboundary displacements. Proper choice of a setof distribution functions(here it is foundto be Legendrepolynomials) for thesebound- 11:05 ary displacementsleads to efficientevaluation of the kernelof the bound- U10. Analysis of finite element •nd finite difference metho•L•for the ary integralby meansof globalmatrix approach, as implementedin the "parabolic"equation. Georgios D. Akrivis and VassiliosA. Dougalis SAFARI code [H. Schmidtand F. B. Jensen,J. Acoust. Soc. Am. 77, (Department of Mathematics, University of Crete, lraklion, Crete, 813-825 (1985)]. This leadsto a matrix relation betweenthe displace- Greece) ment and stressamplitudes along the vertical boundaries,similar to the stiffnessmatrix of finite elementmethod. Once solvedfor the displace- Somerecent results of the authorson theconstruction and analysisof mentamplitudes at the sectorboundaries using a Galerkinapproach, the finiteelement and finite difference methods for theSchr6dinger equation boundaryintegral formulationdirectly yieldsthe total field within each (viewedas one-dimensional "parabolic" approximation to the Helmholtz sectorincluding both the forwardpropagating and reverberatedfield. A equationin underwateracoustics) will be surveyed.The resultsinclude fewexamples of the canonicalproblems will be givento explorethe feasi- the construction,analysis, and implementationof efficient,conservative bility and efficiencyof the proposednew methodfor solvingvarious finiteelement methods of highorder of accuracybased on splineapproxi- range-dependentocean acoustic propagation problems. mationsin the depthvariable and suitablePad6-type techniques in the range.In addition,optimal error estimatesfor classicalsecond-order ac- curatefinite difference techniques of Crank-Nicolsontype for the "para- bolic"equation with interfacesand variable bottom will be presented. 10:35 US. The problem of energy conservationin one-way wave equations. Michael B. Porter, Carlo M. Ferla, and Finn B. Jensen (SACLANT UnderseaResearch Centre, 19026La Spezia,Italy) 11:20 A recentattempt to defineand numericallysolve range-dependent UI 1. A marchingmethod for someelliptic modelsof wave propagation. benchmarkproblems in oceanacoustics [J. Acoust.Soc. Am. Suppl.1 81, George H. Knightly and Donald F. St. Mary (Department of S39-S40 (1987) ] revealedsome inherent problems with energyconserva- Mathematics, University of Massachusetts,Amherst, MA 01002) tionfor one-waywave equations when applied to slopingbottom environ- ments.The problemturned out to beassociated with the useof a stair-step Parabolicequation (PE) methodsfor wavepropagation in the farfield approximationto a slopinginterface, and, in particular, with the rise of are frequentlyobtained from ellipticequations valid in the farfield.Here, the stair steps,where a one-wayequation allows only one out of two propertiesof the ellipticmodels are discusseddirectly. The equationsare necessaryinterface conditions to be fulfiled (continuity of pressureor discretizedand a marchingmethod is obtained.Although the underlying horizontalparticle velocity). The lossof energyincreases with increasing initial valueproblems are not well posed,in certainparameter ranges, the slope and density contrast at the water/bottom interface, and it can be- marchingmethod is stablewhen the stepsizes are restrictedin a suitable comequite significant (3-6 dB) for slopesof a few degreesand a density way. Under suchrestrictions, propagation loss curves are obtainedthat contrastof 2. A simplefix to this problemis to solvefor a density-reduced showgood agreement with exactsolutions of sometest problems.Both pressure(p/,f•) insteadof pressure (p), a trivialmodification toexisting fluid and elasticwave problems are considered.[Work supportedin part one-wayparabolic equation codes. Numerical resultsfor both up- and by ONR. ] S54 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S54 WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM C, 8:35 A.M. TO 12:00 NOON SessionV. ArchitecturalAcoustics III: Acousticsof TeleconferencingFacilities David L. Klcpper,Chairman KlepperMarshall King Associates,Ltd., 7 Holland Avenue,White Plains, New York 10603 Chairman's Introduction4:35 Invited Papers 8:40 V1. Teleconferencing---Theacoustical environment and directional transducers. J. E. Westand G. W. Elko (AcousticsResearch Department, AT&T Bell Laboratories,Murray Hill, NJ 07974) The performanceof teleconfetericing systems are largely a functionof theacoustical environment, the type, andthe location of loudspeakersand microphones. Acoustical treatment in mostconference rooms is limitedto carpetedfloors and an acousticalceiling. For the mostpart, presentmicrophones are designedto work best under closetalking conditionsand have omnidirectional,or first-orderunidirectional characteristics. Two smallexperimental conference rooms have been constructed in which the effectsof room parameterssuch as reverberationtime, noiselevel, and discreteecho control can be independentlystudied. Broadband (8-kHz) directionalmicrophones and steerabledirectional arrays, mostly of secondorder, are beinginvestigated for a varietyof acousticalenvironmental conditions. The easilyadjustable conference rooms allow us to investigate thesemicrophone systems with respectto their directionalproperties, placement, and steeringalgorithms. Loudspeakerdirectional characteristics, and advantages of stereotransmission are alsounder investigation. The experimentalconference rooms and newdirectional microphones will be describedand a discussionof trade-offsbetween major parameters for speechreception and transmissionwill be given. 9:10 V2. Microphonesand feedbacksuppression in teleconferencing.Michael Pettersonand SteveJulstrom (ShureBrothers, Incorporated, 222 Hartrey Avenue,Evanston, 1L 60202-3696) The effectof room acousticson microphoneplacement and selectionfor teleconferencingsystems and currentapproaches to electroacousticfeedback and echo control in teleconferencingsystems will bediscussed. V3. Teleeonfereneeroom acoustics.R. Kring Herbert (OstergaardAssociates, 115 BloomfieldAvenue, Caldwell, NJ 07006) In dedicatedteleconference rooms, there is an acousticalconflict between finishes for optimummicrophone pick-upand those for optimumspeech intelligibility within the space. This paperlooks at the implicationsof this for teleconferenceroom finishesand geometriesas well asother acousticallycritical factorssuch as sound isolationand ambientnoise control. Recommendations are givenfor optimumperformance. 10:10 V4. Room acousticand soundsystem design for full duplex teleconferencefacilities usingceiling-mounted transducers.Wade R. Bray (JaffeAcoustics, Inc., 114-A WaghingtonStreet, Norwalk, CT 06854) Room acousticsand the effectsof differingconditions on performanceof a full-duplexaudio telecon fetenee system,which nulls direct-pathcoupling, leaving only couplingby reflectedenergy, were studied.Room impulseresponses and signal-to-noiseratios at talker locationsand at transducerlocations were taken for roomswith differingacoustic conditions, as werebinaural recordings of the impulseresponse tests and of sampleteleconferences in the samesituations, for subjectivecomparison and simultaneousanalysis. Long workingdistances from talkersto microphonesoffer the advantageto usersof freedomof placementand movement,but at the costof strict limits on ambientnoise loudness and, more particularly,noise structure (annoyancecomponents). Early reflectionswithin the psychoacousticfusion time limit describedby the Haas effectwere found beneficial, while reflections beyond that timeproved detrimental both to intelligibilityand to circuitfunction. Discrete energetic late reflectionsare especiallyundesirable. Optimization for variousroom configurationswas studied,and criteria presentedfor obtainingsubjectively high-quality teleeonferences between rooms and from rooms to handsets. S55 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $55 10:40 VS. Employingthe completeacoustical palette in teleconferencingdesign. Peter D'Antonio (RPG Diffusor Systems,Inc., 12003Wimbleton Street, Largo, MD 20772) The prevailingarchitectural acoustics approach in teleconferencingis "Deader is better" or "No acoustics is betterthan bad acoustics."In somesituations, this is appropriate,but thereare alternativeapproaches to reflectioncontrol in teleconferencingfacilities which make use of the entireacoustical palette, namely absorp- tion,reflection, and diffusion. Since the application of diffusivesurfaces is newto theteleconferencing environ- mcnt,a reviewof the developmentand the spatial,temporal, and spectralproperties will bc presented.A teleconferencingdesign for a hoardroom will bcproposed that makesuse of an integratedsuspended ceiling cloud,containing the variouselements of the acousticalpalette, as well aslighting, and HVAC elements.This designprovides diffuse sound across the conference table, which allows a comfortableconversation level, better acousticcoverage and communication among the participants, as well as a morenatural transmitted sound to the otherend of the converstation.The ceilingcloud also affords a convenientapproach for concealingthe necessaryHVAC ducts,wiring, etc. and alsoprovides accessibility for maintenance. 11:10-11:15 Break ContributedPapers 11:15 11:30 V7. Adaptiveacoustic echo cancellation algorithms in teleconferencing V6. An acousticecho cancelerimplemented with AT&T DSP16A. Sen systems.Sen M. Kuo and Huan Zhao (Department of Electrical M. Kuo and Huan Zhao (Department of Electrical Engineering, Engineering,Northern Illinois University, DeKalb, IL 60115) Northern Illinois University,DeKalb, IL 60115) In teleconferencingsystems, acoustic echoes consist of acousticcou- plingbetween louder speakers and microphones and the roomreverbera- Acousticecho control in teleconferencingand hand-freephone sys- tions.The echoescould be very annoyingand evencause howling if the temsis not trivial becauseof its specialcharacteristics. Many acoustic echolevel is high enough. Among many ways of canceling acoustic echoes cancellationalgorithms have been published in recentyears but they are in teleconferencingsystem, the adaptivedigital signalprocessing ap- eitherof theoreticalconsideration or arecomputer simulations with com- proachis very promising since adaptive filtering has the capability to trace plicatedcomputation burdens, which makesreal-time implementation the changeof the echopath and cancelechoes actively. Adaptive algo- impossible.In this paper,a two-bandscheme is proposedto realizean rithms to cancelacoustic echoes in both the time and frequencydomains acousticecho canceler with an AT&T WE DSP 16A microprocessor.The are presented,and the computersimulation results are comparedin a structureof the algorithmis optimizedso it cangive an echoreturn loss proposedpaper. The algorithmsincluded are LMS, NLMS, LLMS, enhancement(ERLE) of more than 25 dB. A double talk detector is SDLMS, SELMS, SSLMS for FIR adaptive filtering and RLMS, developedwith the subbandadaptive algorithm so that it will ensurefull SHARF, andPole-Zero for IIR adaptivefiltering. Performance of these duplexcommunication. Performance of this canceleris evaluatedwith adaptationalgorithms will bediscussed. Subband adaptive filtering algo- both computersimulation results and real-timetest results. rithm, whichhas a fasterconvergence rate, will alsobe presented. 11:45-12:00 OpenDiscussion S56 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America $56 WEDNESDAY MORNING, 29 NOVEMBER 1989 DIRECTORS' ROW 42, 9:00 A.M. TO 12 NOON Meeting of Accredited Standards Committee S2 on Mechanical Shock and Vibration to be heldjointly with the TechnicalAdvisory Group (TAG) Meeting for ISOfYC 108 MechanicalVibration and Shock S. I. Hayek, ChairmanS2 AppliedResearch Laboratory, Pennsylvania State University, P.O. Box30, StateCollege, Pennsylvania 16801 D. F. Muster, Chairman,Technical Advisory Group for ISO/TC 108 4615 O'Meara Drive, Houston, Texas 77035 StandardsCommittee S2 on MechanicalShock and Vibration.Working group chairs will presentreports of theirrecent progress on writing and processing various shock and vibration standards. There will bea reporton the interfaceof S2 activitieswith thoseof ISOfrC 108 (the TechnicalAdvisory Group for ISOfrC 108 consistsof membersof S2, S3, and other persons not necessarily members of thosecommittees) including plans for the nextthe meetingof ISO/TC 108,to be heldin Milan, Italy from 26 Marchto 6 April 1990. WEDNESDAY MORNING, 29 NOVEMBER 1989 ST. LOUIS BALLROOM A, 9:00 A.M. TO 12:30 P.M. SessionW. Musical AcousticsII and PsychologicalAcoustics IV: Timbre Ira J. Hirsh, Chairman WashingtonUniversity and CentralInstitute for theDeaf, St. Louis,Missouri 63130 Chairman's Introduction--9:00 Invited Papers 9:05 Wl. Theperceptionoftimbreofstcady-statecomplextones.Reinier Plomp(DepartmentofOtolaryngology, FreeUniversity Hospital, P.O. Box7057, 1007MB Amsterdam,The Netherlands) The paperwill presenta reviewof workby theauthor in timbreperception. Steady-state complex tones are characterizedby threeperceptual attributes: loudness, pitch, and timbre.Loudness and pitch are mainly determinedby a singlephysical parameter: loudness by intensityand pitchby the fundamentalfrequency. Timbre,on the contrary, is not a one-dimensionalattribute: It isprimarily determined by the shape of thesound spectrumand, therefore, multidimensional. This aspect of tonesensation was investigated bymeans of multidi- mensionalscaling_ For tonesderived from musicalinstruments, excellent agreement between three-dimension- al timbrerepresentations and three-dimensional spectrum representations was found. The mostimportant dimensionis the slopeof the spectrum:tones with stronglow harmonicssound dull, toneswith stronghigh harmonicssound sharp. Local peaks in thespectrum result in specifictimbres as in speechvowels. It wasfound that the phaserelation among the harmonicshas only a minoreffect on timbre.The importanceof the latter findingin relationto the effectof reverberationon timbrewill alsobe discussed. S57 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S57 9:35 W2. Timbre and tone height. Roy D. Patterson (MRC Applied PsychologyUnit, ]5 Chaucer Road, CambridgeCB2 2EF, England) A soundcomposed of all harmonicsof 131Hz in cosinephase produces the noteC3 with a buzzytimbre. Onecan produce similar notes that are all C's but that sound"higher" than the originalC, in at leastthree differentways. One can (a) attenuatethe lower harmonics,(b) attenuatethe odd harmonics,and (c) shift the phaseof eitherthe odd or the evenharmonics. Traditionally, these perceptual changes would be regardedas timbrechanges because tone chroma is fixed,but thesemanipulations also affect the toneheight of the sound. Accordingly,experiments were performed to investigatethe relationshipbetween timbre and toneheight. The stimuliwere presented at sixoctaves with fundamentalsranging from 32-1024 Hz. The listenersjudged the octaveof thesounds on a scalefrom 1 to 6. The resultsshow that waveswith thesame period can produce tone- heightjudgments that consistentlydiffer by morethan an octave,and that a substantialcomponent of many presumedtimbre differences (e.g., betweena pianoand a harpsichord)is actuallya tone-heightdifference. 10:05 W3. Pitch and timbre interactionin isolatedtones and in sequences.Paul Iverson and Carol Krumhansl (Departmentof Psychology,Uris Hall, CornellUniversity, Ithaca, NY 14853-7601) The interactionof pitchand timbreof tonesin isolationand in sequenceswas studied in a seriesof experi- ments.Experiment I investigatedwhether pitch, timbre,and durationinteract in the detectionof changes betweentone sequences. It was found that variations in timbreinterfered with thedetection of pitchchanges, as did the variationof pitchwhen detecting timbre changes. Changing durations did not interferewith detection of eithertimbre or pitchchanges. Experiment 2 investigatedwhether the interactionbetween timbre and pitch is a consequenceof changesof relationsbetween tones in the sequences.Contextual variation between se- quencesin pitchinterfered with pitchjudgments, but no othercontextual interference was found. Experiment 3 measuredreaction times of variousclassifications of isolatedtones varying in pitch and timbre and found that pitchand timbreinteract. These results indicate that pitchand timbreinteract at the levelof individualtones, theydo notinteract at a structurallevel, pitch is coded relationally, and timbre is not coded relationally. [ Work supportedin part by a Grant-in-Aid of Researchfrom SigmaXi, The ScientificResearch Society. ] 10:35 W4. Interactionof timbreand pitch in spectraldiscrimination tasks using complex tones. Punita G. Singh (Central Institute for the Deaf, St. Louis, MO 63110) The dualnature of pitch("spectral"/"virtual") and its relationto suchtimbral percepts as "sharpness" and"roughness" isexplored in a setof experimentsdesigned to ascertainperceptual cues facilitating discrimi- nationtasks with complextones having fiat spectralenvelopes. Three types of complexesare employed: ( 1 ) harmonic"residue" tones comprising four harmonics, (2) ten-componentharmonic and inharmonic complex- eswith all componentsshifted from some reference frequency, and (3) ten-componentcomplexes with a single componentshifted from its harmonicfrequency. Using two-tone sequences as stimuli, listeners are askedto judgeif thesecond tone is (1) thesame, (2) higherin pitch,(3) lowerin pitch,(4) differentin "something else,"(5) differentin "somethingelse" and higher, or (6) diffferentin "somethingelse" and lower in pitch than the first ("somethingelse" is takento be synonymouswith "timbre"). For residuetones, data indicate thatchanges in spectrallocus yield changes in timbre.Further, direction of locuschange can indicate pitch change,despite little or no changein F0 (0%-2% for 200- and 400-Hz F0). This impliesthat changesin timbrai"sharpness" may be construedas pitch changes, given the absenceof othercues. For stimuliin which componentsare shifted from harmonicfrequencies, the unitarysensation of a complexmay be replaced by one of multiple sources.Low components(n = 1-3) are more susceptibleto being "heard out" as individual entities,while changes in highercomponents may yield changes in timbre,such as "roughness." Thus compari- sonsof componentsbased on factorssuch as ( 1) magnitude,(2) location,and ( 3 ) directionof spectralchange in a sequencecan guidegrouping operations, in additionto indicatingchanges in an overall propertyof a complexas a whole,such as its pitch,timbre, or bothof thesepercepts simultaneously. [Work supportedby AFOSR. ] 11:05 WS. Timbre and the perceptual organization of musical patterns. David Wessel (Center for New Music and Audio Technologies,Department of Music, Universityof California,Berkeley, Berkeley, CA 94720) Timbre, usuallydefined as the conditionof attributesother than pitch, loudness,and duration, playsa strongrole in determiningthe perceptualorganization of musicalpatterns. Timbre's primary organizational influenceappears to be on perceptualgrouping, as in auditorystream segregation and rhythmicsegmentation. Groupingby timbre can influencethe tonal implicationsof otherwiseambiguous pitch material.Extensions of the author's previouswork on the perceptionof timbral analogiesdemonstrate that timbre melodiesand timbrallybased phrasing and articulationcan exhibit invariances similar in characterto the mannerin which pitch patternsexhibit perceptual invariance under transposition.With the strongorganizing influence of S58 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S58 timbrein mind,it isargued that care should be taken in experimentson melodic and rhythmic organization to controlfor theinfluence of timbre.Pure tones and fixed spectra exhibit a strongcorrelation between fundamen- tal frequencyand measures of thelocation of thespectral envelope. an important timbrai determinant and are inappropriatefor manyexperiments on musicperception. A simpleset of proceduresfor preparii•gstimuli with independentcontrol of spectralenvelope, temporal envelope, and frequency content is providedand demon- strated. ContributedPaper 11:35 tonesused in an earlierstudy [$. Grey, J. Aeoust.Soc. Am. 61, 1270-1277 (1975) ]. FollowingGrey's three-dimensionalmodel, the data were ana- lyzedfor theeffects of harmonicsynchrony, spectrum content (eentroid), W6. Effect of spectrumand attack propertieson the evaluation of and attackcharacteristics. The latter two attributesshowed a significant concurrently sounding timbres. Grego• Sandell (Northwestern relationshipto blend: in general,"dark" instrumentsand instruments Computer Music, Northwestern University, 2299 Sheridan Road, with quick, clean attacks tended to blend well. The amount of blend re- Evanston, IL 60208 } portedfor suchinstruments decreased as their pairsincreased in bright- hessor attacklength, and pairs of brightor noisy-attackinstruments did The perceptionof concurrentlysounding musical timbres underlies not blend well. The results found here corroborate some of the recommen- musicalorchestration, but currentknowledge about psychological attri- dations for desirable blends in traditional orchestration manuals. Other butesof timbreemphasizes successive comparison (e.g., similarity, analo- aspectslikely to pertainto orchestrationare interactions between sounds gy). Thepsychological attribute "blend" was investigated by havingeight (e.g., masking,acoustic consonance) and the differentialsalience of listenersuse a ten-pointscale (a continuumof"separated" to "fused")to timbres(e.g., recognizability, loudness, stream segregation ); thesewill be rate all possiblesimultaneous presentations of synthesizedinstrumental evaluated. 11:50 Open Discussion $$9 J. Acoust.Soc. Am. Suppl.1, VoL86, Fall 1989 118th Meeting:Acoustical Society of America S59 WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM F, 1:00 TO 2:40 P.M. SessionX. EngineeringAcoustics lI and Noise IlI: AcousticMetrology: Measurements,Calibration, and Standards GeorgeS. K. Wong, Chairman Laboratoryfor BasicStandards, Division of Physics,National ResearchCouncil, Ottawa, Ontario K1A OR6, Canada Invited Papers 1:00 Xl. Calibration of underwater acoustic transducers at NRL/USRD. A. L. Van Buren and J. E. Blue (UnderwaterSound Reference Detachment, Naval ResearchLaboratory, P.O. Box 8337,Orlando, FL 32856- 8337) At the beginningof World War II, the United Statesrecognized the needfor establishingsystematic methodsfor calibratingand evaluatingsonar transducers and establishedthe UnderwaterSound Reference Laboratory.The UnderwaterSound Reference Detachment of the Naval ResearchLaboratory evolved from thisbeginning and now serves as the focus for underwateracoustic metrology in theUnited States. This paper coversthe measurementfacilities and methodsfor carryingout the metrologyfunction. Evaluation of sonar transducersand someof the ongoingresearch in calibrationmeasurement methodology are alsodiscussed. Finally,some comments are made about calibration system requirements for implementingnew measurement techniques. 1:25 X2. Precision electroacousticalmeasurements at the NIST (formerly the NBS): Selected examples. Victor Nedzeinitsky(National Institute of Standardsand Technology, Sound Building (233), RoomAI47, Gaithersburg,MD 20899-0001) In additionto primarycalibrations determining the pressure and free-field sensitivity lev. els of laboratory standardmicrophones by reciprocitytechniques, a broad variety of electroacousticalmeasurements are per- formedat the NIST. Theseinclude the calibrationand characterization of measuringmicrophone systems, audiometersand audiometric carphones, and other instruments at verylow sound-pressure levels, as well as the investigationof the linearity of theseinstruments over relatively large dynamic ranges. Techniques and appara- tus usedfor thesepurposes include free-field measurements in the NIST anechoicchambers and sound- pressuremeasurements in acoustic couplers. Separation of signalsfrom noise is achievedby meansincluding analogfilters, synchronous detectors (lock-in amplifiers), and dynamic signal analyzers. These incorporate a varietyof excitationsource types (e.g., sine-wave, chirp, bandlimited-impulse, periodic noise, and random noisesignals ) and signal-processing techniques (e.g., FI•T, extractionofthe fundamental component of a sine- wavesignal from a digitizedtime record). Specific examples from the NIST researchand measurement services arc discussccl. 1:50 X3. Aeoustiealinstrument calibration and standards. George S. K. Wong (Laboratoryfor BasicStandards, Divisionof Physics,National ResearchCouncil, Ottawa, Ontario K 1A 0R6, Canada) In the field of acousticalmetrology, one oftenencounters questions such as: Why do we needto calibrate instrumentsfrom timeto time and at whatintervals? How doesone select an acceptablecalibration procedure to implementtraceability with a specificuncertainty? In orderto haveconfidence in measurements,the overall measuringsystem should be calibratedin accordancewith the requirementsof a standardor an accepted procedure.In general,"type testing"refers to teststhat are essentialto verify the designintegrity of the instruments;"calibration" refers to teststhat are designedto monitoroperational performances which may changewith the passageof time;and "fieldcheek" is a relativelysimple procedure, which can be performed in situ,to givesome indication of whetherthe instrumentis operationaland functioningproperly. To implement traceabilityfor a specificuncertainty of a calibrationprocedure is a completelydifferent matter. Examples and factors that determine the calibration schedule are discussed. S60 J. Acoust.Soc. Am. Suppl.1, VoL86, Fall 1989 118th Meeting:Acoustical Society of America $60 2:15 X4. Acousticcalibration requirements in an aerospaceindustry environment. L. Kasturi Ranganand KlausJaeger (Lockheed Missiles and Space Company, Inc., 0/4875, B/195A, Sunnyvale, CA 94088-3504) Severaltypes of acousticmeasuring instruments are usedat LockheedMissiles and Space Company's varioustest facilities. Lockheed's Primary Standards Laboratory is required to calibratethese instruments. In thispaper, the environmentswhere the instrumentsare used,calibration requirements, traceability require- mentsand howthey are met, calibrationrecall intervals, and uncertaintiesin thecalibration standards hierar- chy will be presented. WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM G, 1:00 TO 3:30 P.M. SessionY. Physical AcousticsIV and EngineeringAcoustics III: Localized Wave Modes Bill D. Cook, Chairman Departmentof MechanicalEngineering, University of Houston,Houston, Texas 77204-4792 Chairman's Introduction--l:00 Invited Papers 1:05 ¾1. Localizedwave transmission physics and engineering.Richard W. Ziolkowksi(Engineering Research Division,P.O. Box 5504, L-156, LawrenceLivermore National Laboratory,Livermore, CA 94550) Exactsolutions of the scalarwave equation that describelocalized transmission of waveenergy will be reviewedbriefly. These acoustic solutions can be optimizedso that they are localizednear the directionof propagationand their original amplitude is recoveredout to extremelylarge distances from their initial loca- tion. Pulseswith thesevery desirable localized wave transmission (LWT) characteristicsmay have a number of potentialapplications in the areasof directedenergy applications, secure communications, and remote sensing.The followingwill beshown. ( 1) The underlyingphysics of the LWT effectis closely connected to the additionaldegree of freedomobtained by couplingthe usuallydisjoint portions of phasespace. Position and frequencyspectra at differentlocations in the apertureare differentbut are highlycorrelated so that the effectivefrequency of theaperture is higher than expected and the resulting pulses from these aperture sources reconsitutethe frequencycontent of the LWT packet,hence its pulseshape, as it propagates.(2) The LWT solutionsdo not violateany known uncertainty relations and satisfy a generalizedfull phase-spaceuncertainty relation.(3) The independentlyaddressible element LWT arrayscan be designed to dramaticallyoutperform conventionalcw-driven apertures. This includespulse shape, peak amplitude, and energy fluenee. New acous- tic experimentaldata that corroboratethese theoretical results will alsobe presented.[This work wasper- formedby theLawrence Livermore National Laboratory under the auspices of theU.S. Department of Energy under Contract No. W-7405-ENG-48. ] 1:25 ¾2.NondiffraetingBesseloptlcs:Theory and practice. J. H. Eberly (Departmentof PhysicsandAstronomy, Universityof Rochester,Rochester, NY 14627) The theoreticalfoundation for the nondiffractingBessel beams identified in 1985 by Durnin will be ex- plainedbriefly, and severalpractical realizations of thesebeams in opticalcontexts will be described. 1:45 ¾3. Experimentalevidence of localizedwave phenomena. D. Kent Lewis,Steven E. Benson,Floyd C. Kirk (MechanicalEngineering Department, Lawrence Livermore National Laboratory,Livermore, CA 94550), andBill D. Cook (CullenCollege of Engineering•University of Houston,Houston, TX 77204-4792) Recentinterest in creatingbeams that exhibitmore localizedbehavior on the propagationaxis than tradi- tional solutionshas led to many formalismsfor creatingand describingenergy transport. One set of these S61 J. Acoust.Sac. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of Amedca $61 solutions,the localizedlinear wave solution developed by Ziolkowski,has been under investigation through acousticexperiments for sometime. Experimentsusing linear superposition to simulatetwo-dimensional acousticarrays are discussed. These experiments use two methods, acousto-optics, to simulate an arrayof line sourcesand a pointdetector, as well as the traditional two-transducer arrangement, to simulatesquare arrays of pointsources. Preliminary experiments with an actualtwo-dimensional acoustic array are alsodescribed. The techniqueof signalpreprocessing to overcome damping problems and interelement interactions is dis- cussed.[Work performedby the LawrenceLivermore National Laboratory under the auspicesof the U.S. Departmentof Energyunder Contract No. W-7405-ENG-48.] 2:05 Y4. Propagationof localizedpulse trains in a circularacoustic waveguide. I. M. Besieris,A.M. Shaarawi (Bradley Departmentof Electrical Engineering,Virginia PolytechnicInstitute and State University, Blacksburg,VA 24061), and R. W. Ziolkowski(Engineering Research Division, LawrenceLivermore NationalLaboratory, P.O. Box 808, L-156, Livermore,CA 94550) A briefoverview of a novelapproach to the synthesis of wavesignals [cf. I. M. Besicris,A.M. Shaarawi,and R. W. Ziolkow.ski,J. Math. Phys.30, 1254(1989)] will be presented.This approach, referred to as the bidirectionalmethod, was originally introduced in orderto understandthe salient features of Brittingham-like solutions.Its scopeis broader, however, and encompasses classes of problemsaltogether different from wave propagationin an unboundedhomogeneous domain. The efficacyof the bidirectionalmethod in geometrics involvingboundaries has alreadybeen demonstrated [cf. A.M. Shaarawi,I. M. Besieris,and R. W. Ziol- kowski,J. Appl. Phys.65, 805 (1989)]. In thispresentation, the propagationof localizedpulse trains in an infinitelylong, circular, acoustic waveguide will be examinedin detail.The farfieldsradiated out of a semi- infinite,circular, acoustic waveguide, excited by a localizedinitial pulse,will alsobe studied.These approxi- matesolutions, which are computedusing KirchhoWs integral formula with a retardedGreen's function, are causal,have finite energy, and exhibita slowenergy decay. 2:25 YS. Pulsedfocused fields. Leopold B. Felsen(Department of ElectricalEngineering and ComputerScience, PolytechnicUniversity, Farmingdale, NY 11735) Highlyconcentrated clumps of fieldenergy that movethrough an environmentwithout diffusing are of interestfor a varietyof applications.The basicproblem is the synthesisof sourcedistributions over an initial apertureto achieve this objective. Various analytical approaches have been used to explore how initial transient sourcefield distributions can be tuned up soas to generatecompact wavepackets that staytogether indefinitely. Themeaning of "stayingtogether" and "indefinitely" depends strongly on how the problem is defined and has led to somestartling conclusions in the early phasesof investigation.At present,there seems to be agreement that no radicallynew phenomenaare operativehere but that interestingpulse shapes can be syntheziedby cleverspectral tuning of inputconditions. Various techniques--based on directtime-domain synthesis, real andcomplex spectral synthesis, and use of transientbeam-type basis elements--are reviewed within the con- textof causalityand finite aperture size. Also discussed isa modelingscheme whereby a pulsedfocused beam is generatedanalytically by assigning complex values to the source location and initiation time. This parametriza- tion is usefulfor convertingspherical pulse interaction with a perturbstingenvironment directly into focused pulseinteraction with that environment.[Work wassupported by the InnovativeScience and Technology Officethrough the U.S. Army Harry DiamondLaboratory.] Contributed Papers 2:45 58, 1499 (1987) ] showedthat a Besselbeam of light wasdiffraction-free. Usinga noveltechnique of nonuniformpoling, the polarizationstrength Y6. An experimentalstudy of the field profileof a Besselbeam ultrasonic of a piezoelectricceramic disk was made to followthe pattern of anaxially transducer.D. K. Hsu, F. J. Margetan, and D. O. Thompson (Ames symmetric,truncated Bessel function J• in both amplitudeand phase. Laboratory,Iowa StateUniversity, Ames, IA 50011) Field profilesand propagationbehavior of the Besselultrasonic beam The firstexperimental investigation of ultrasonicbeams produced by a weremeasured experimentally in a waterimmersion tank. The measured Besseltransducer is reported.Bessel sources are of particularinterest in profilesagreed well with calculatedresults using a Gauss-Hermitebeam attemptsto achievediffractionless beams; Durnin et aL [ Phys.Rev. Lett. model.Effects of the numberof lobes,frequency, and beamwidth on $62 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S62 diffractionbehavior were investigated. [Work supported by Basic Energy 3:15 Sciences,U.S. Department of Energy.] Y8. Modificationo[ the NPE computercode to describethe propagation of axisymmetricsound beams in infinlte media.Gee-Pinn James Too and JerryH. Ginsberg(SchoolofMechanicalEngineering, Georgialnstitute of Technology,Atlanta, GA 30332) NPE is a nonlinearprogressive wave equation computer code devel- oped by McDonald and Kuperman [Comp. Math. Appl. 11, 843-851 (1985) ] to evaluatetransient propagation in acousticwaveguides. It is 3..OO suitablefor two-dimensionalphenomena, as well as for radialpropagation Y7. Developmentof an acousticarray for generatingADEPT pulsesin of azimuthallysymmetric waves. The presentstudy describes the modifi- air. Alan P. Poindcxter, Bruce D. Baker, and Bill D. Cook (Cullen cationsrequired to employNPE for the evaluationof axially propagation Collegeof Engineering,University of Houston,Houston, TX 77204- axisymmetricwaves, particularly those associated with sound beams radi- 4792) ated by a baffledpiston. In additionto implementinga formulationin whichthe propagation is essentially parallel to theaxis of a setof cylindri- A six-ringcircular array with eachring driven by separateclectrnnics cal coordinates,it is necessaryto accountfor the transversespreading of hasbeen developed for the purposesof exploringacoustic directed energy the beam into an infinite medium. Another issue is the manner in which pulsetrains in air. Eacharray clement is a smallloudspeaker with 1-in.- theinitial waveform input to NPE isobtained. The predictionsof NPE for diamcone. A differentsignal is to beapplied to eachring simultaneously. a linearsound beam when the inputis obtainedfrom the King integral, Thesesignals come from an MACII computerwith an off-the-sbelfinter- whichis an exactsolution in quadratureform, are compared to theresults faceboard. The signalmust be compensated for the 1oudspeaker'ssystem obtainedfor a simpleinput based on assumptionof planarwave behavior reponseand interaction between array elements. [This work is supported in the vicinity of the transducer.The resultsin both casesare alsocom- by LawrenceLivermore National Laboratory.] paredto the analyticalsolution for the farfieldradiation pattern. WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 DIRECTORS' ROW 43, 1:00 TO 2:50 P.M. SessionZ. Structural Acousticsand Vibration III: Damping in Structural Systems Martin J. Pechersky,Chairman AppliedResearch Laboratory and Departmentof EngineeringScience and Mechanic•Pennsylvania State University, UniversityPark, Pennsylvania16802 Chadman's Introduction•l:00 ContributedPapers 1:05 thoughthe effect of blendingon the physical properties of neopreneMends hasbeen investigated, the effectof blendingon the dynamicmechanical Z1. Dynamic modulusmeasurements made with the dynamic mechanical propertieshas not been investigated to thesame extent. Neoprene is com- thermal analyzer. R. J. Deigan and J. J. Dlubac (Code 1965, Ship patiblewith naturalrubber, SBR, NBR, polysulfiderubber, and has limit- AcousticsDeparlment, David Taylor ResearchCenter, Bethesda,MD ed compatibilitywith bulyl rubber.Seventy different blends were pre- 20084-5000) paredand the physicalproperties measured. The dynamicalmechanical Polymer Laboratories' dynamic mechanical thermal analyzer propertiesare beingmeasured on a PolymerLaboratories DMTA and ( DTMA } measuresthe dynamicYoung's or shearmodulus and lossfac- frequencyshifted using the WLF equation.This paper presents data and tor of viscoelastie materials. The DMTA infers the modulus and loss an analysisof the samplesmeasured thus far. factorby measuringthe responseof a smallsample to forcedvibrations in a small material sample.An optimal samplegeomelry is selectedfor a given Young's modulusrange by consideringeffects such as thermal stress,shear deflection in shortbeams, and system compliance. The goalis 1:35 to minimize thesecombined errors in the dynamic Young's modulusand lossfactor measurement over the full madulusrange. As a resultof this Z3. Structuraldamping measurements of polymercomposites with laser analysisrecommended DMTA test proceduresare presented. Doppler vibrometry.Martin J. Pecherskyand Michael T. Gemsky (AppliedResearch Laboratory, The PennsylvaniaState University, P.O. Box 30, StateCollege, PA 16804) Dampingloss factors of threecommercially available composites up 1:20 to a frequencyof 12kHz havebeen performed. A resonancedwell tech- z2. Dynamic mechanical properties of neoprene rubber blends. W. nique was usedin which both the excitationand measurementwere non- M. Madigosky(Naval SurfaceWarfare Center, 10901New Hampshire contact.The excitationwas accomplishedelectromagnetically and the Avenue,Silver Spring, MD 20903) drivingpoint admittance was measured with a laservibrometer. The 25- era-longslender beam specimens were seismically supported at two nodes Blendsof neoprenewith otherelastomers can be usedto providecer- for eachresonance by a combinationof softsprings and silk thread.The tain propertiesthat are not obtainablewith either elastomeralone. AI- modeshapes to determinethe nodesof eachbeam were measured prior to $63 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S63 the dampingmeasurements. Metallic specimens were also measured to the fluid,the equationof lossfactor of the wholesystem was derived. The determinethe limit on the accuracyof the measurement.The damping analyticalresults showed a very goodqualitative agreement with the ex- factorsfor thecomposites (Q-•) werefound to beabout two orders of periment.The study showedthat different shell modesfiad a different magnitudelarger than the metallicbeam and all of the compositedamping contributionto the squeezingfilm damping,and theoptimal design of the factorswere found to increasemortatomically with theexcitation frequen- doublepipe dampingsystem can be deduced.The theory has beenex- cy overthe rangetested. Loss factors of the compositesranged from I to tendedto studyingthe squeezing film damping system with variouskinds 4X 10-3 at thelower frequencies to about 3 to 16X 10-3 at thehigher of viscousfluid, suchas lubricationoil, etc. frequencies. 1:50 2:20 Z4. Dynamic finite element analysis of a compressionallyloaded 7.6. Obtaining internal and coupling loss factors by a transient test viseoelasticcylinder. R. W. Harrison (Naval SurfaceWarfare Center, method. K. Sato and N. Lalor (Institute of Sound and Vibration SilverSpring, MD 20903-5000) Research,University of Southampton,Southampton SO9 5NH, United A nonlinear finite element program utilizing the Mooney-Rivlin Kingdom) strain-energytensor was used to obtainthe static deflection of a compres- Statisticalenergy analysis (SEA) is takingan increasingrole in deter- sionallyloaded viscoelastic cylinder. The deformedcoordinates and the miningenergy flow and predicting noise reduction of machinestructures. massand stresstensors were then passedto a complementarydynamic In orderto setup the SEA modelit is necessaryto obtainthe internaland finiteelement program where the vibrationisolation as a functionof the couplingloss factors of eachsubstructure either from theoryor from mea- appliedvibrational frequency and the impedanceof the top surfacewere surement.The conventionalmeasurement technique, in which steady- calculated.The predictionsof the finiteelement programs were verified stateinput power is supplied by a shakerto eachsubsystem in turn, is very with both Instron and accelerometer data. The calculations show the ef- timeconsuming. Moreover, it isnot always possible to attachthe shaker in fect of compressionalloading on the vibrationabsorption response and the bestposition due to spacelimitations. This paperdescribes a transient indicatepossible geometrical modifications to enhancethis absorption for method of lossfactor measurement,using an instrumentedhammer, applicationin the theoryof motormount isolation. whichovercomes the above-mentioned difficulties. The techniqueuses the point and transferinertance data for the structureto calculateinput powersand subsystemenergies. Results obtained by this methodare com- 2:05 paredwith thoseobtained conventionally and good agreeement is demon- strated. ZS. Prediction of the loss factor of a double pipe damping system. Meng Li and Robert G. White (Institute of Sound and Vibration Research,University of Southampton,Southampton SO9 5NH, United Kingdom) It is well knownthat increasingthe dampingof machinerystructures 2:35 will reducenoise radiation due to ringingvibration. Damping in industrial Z7. Theoretical and experimental investigation of high-temperature pipescan be increasedby variousdamping techniques. This studywas damping treatment using ceramic-fiber interlayer between plates. L. motivatedby the need to decreasethe noiseradiation and vibration of C. Chow and R. J. Pinnington (Institute of Sound and Vibration pipeworkin powerplants, particularly at elevatedtemperatures. Based Research,The University of Southampton,Southampton SO9 5NH, upon the physicalphenomenon of the air film dampingof two parallel United Kingdom) plates,a doublepipe damping system with a verysmall air gapwas investi- gated.In analysis,the Fluggeshell equations of motionand the Navier- The dampingtechniques available for high-temperatureapplications, Stokesequation for viscousfluid wereemployed. First, a travelingwave- suchas in powergenerating plants, are very limited. As partof theinvesti- typesolution was taken for shelland fluid. Then, from the interfacecondi- gationof high-temperaturedamping work being carried out in ISVR, this tions between the shell and fluid, the solution for the fluid medium was paperdescribes the theoreticaland experimentalassessment of damping expressedin termsof shelldisplacements. Finally, solvingthe fluid equa- of platesusing ceramic-fiber material as an interlayer.An experimental tionof motiongave the fluidvelocity profiles and stresses in theclearance rig that is usedto measurethe ceramic-fibermaterial dynamic properties betweentwo cylindricalshells. From thedefinition of energydissipated in at elevatedtemperature will alsobe described. WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM H, 1:00 TO 3:00 P.M. SessionAA. Underwater AcousticsIV: Computational AcousticsII Henrik Schmidt, Chairman Departmentof OceanEngineering, Massachusetts Institute of Technology,Cambridge, Massachusetts 02139 Cont•buted Papem 1:00 proximationthat adaptivelyadjust range-step size [J. Acoust.Soc. Am. 86, 252-260 (1989) ] and depth-stepsize [Proc. SecondIMACS Symp. AAI. An adaptivetwo-dimensional algorithm for solvingfinite-difference Cornput.Acoust. (to appear)]. Here, an algorithmis presentedthat inte- implementationsof parabolic equations.J. S. Robertson, D.C. Arney gratesprevious work, i.e., both range- and depth-step sizes are adaptively (Departmentof Mathematics,U.S.M.A., WestPoint, NY 10996-1786), selectedby a procedurewithin the implicitfinite-difference (IFD) imple- M. J. Jacobson,and W. L. Siegmann(Department of Mathematical mentationof the parabolicapproximation. An error indicatorthat in- Sciences,Rensselaer Polytechnic Institute, Troy, NY 12180-3590) cludescomponents resulting from both range and depth discretization is In previouswork algorithmshave been developed for enhancingthe computedat eachrange step, and its value is compared to a user-specified efficientsolution of finite-differenceimplementations of the parabolicap- error tolerance window. If the error indicator falls outside this window, S64 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S64 therange- and depth-step sizes are adjusted so that the error is equidistri- of varioustypes provided that appropriateboundary conditions can be butedbetween both components. Further adjustments are not madeuntil expressed.The conditions are incorporated in an implicit finite difference the indicatorleaves the tolerancewindow. In this sense,the algorithm (IFD) schemefor theparabolic equation. The integralsappearing in the generatesa two-dimensional grid that is optimaland that often leads to computingscheme are evaluatedeither analytically, if possible,or nu- substantialdecreases in run time.Examples are presentedthat illustrate merically.For a seabottom considered as an elastichomogeneous half- theaccuracy and efficiency of thealgorithm. [Work supported by ONR. ] space,an impedanceboundary condition can be evaluatedrelating the acoustic field in the sea at the water•ottom interface with that of the elasticbottom. The conditionis insertedin the IFD schemeand the perti- nentintegrals are evaluated numerically by means of a fastFourier trans- 1:15 form algorithm.Results demonslrating the satisfactorybehavior of the schemeare presented for range-independent environments. However, the AA2. Abstract withdrawn. methodis also directly applicable to range-dependentenvironments. 2:00 AAS. Full-wave bottom reverberation modeling. F. Tappert and F. Ryan (Naval Ocean SystemsCenter, Code 541, San Diego, CA 92152) Scatteringby roughnessof thepenetrable water-sediment interface is regardedas volume scattering and two coupledparabolic equations for outgoingand incomingwaves are derived.These equations are solved usinga multiple-forward-scatter/single-back-scatterapproximation plus reciprocity.In thisway, a novelfull-wave, fully range-dependent, bottom reverberationmodel has been developed for long-rangesub-kilohertz fre- quencyapplications. In contrastto all knownexisting reverberation mod- els,scattering strength is not input as a functionof verticalangle. Instead, a statisticalmodel of seafloorroughness is usedin whichthe local power spectrumis inputtogether with range-dependentgeophysical properties of thesediments and basement. The outputis reverberation pressure level (RPL) asa functionof range,bearing, and frequency. Examples of pre- 1:30 dictedRPL in the CanaryBasin operational area will be shown.{Work supportedby ONT. I AA3. Computationalaspects of syntheticseismograms for layeredmedia. P. Getstuft,'} O Vilmann(Odegaard & Danneskiold-SamsoeApS, KroghsgadeI, 2100 Copenhagen,Denmark), and S. Krenk (Aalborg University,9000 Aalborg, Denmark) 2:15 A modifieddirect global matrix methodfor computationof synthetic seismogramsin a layeredviseoelastic media has been developed by com- AA6. Manifestly reciprocal scatteringamplitudes for rough interface bininginformation obtained from continuousand discreteformulations. scattering.David H. Bermanand Dalcio K. Dacol (US Naval Research The resultingcode is stablefor verticalseismic profiling as well asnormal Laboratory,Code 5163, Washington, De 20375-5000) horizontalprofiling and has proven efficient in thegeneration of full-scale seismograms.The wave field is first computedin the wave slownessdo- A newexact expression for scatteringamplitudes for roughinterface main.A properintegration back to the frequencydomain is a criticalpart scatteringis presented.This expressionis explicitlyreciprocal and it is of a forward modelingcode. By consideringthe discretewavenumber shownto hold for a varietyof boundaryconditions: Diriehlet (pressure formulationit is shownthat numericalintegration with equally spaced release),Neumann (rigid surface), impedance,fluid-fluid, and fluid- integrationpoints corresponds to a fictitiouscylindrical reflective surface solidinterfaces. This expression isshown to bea convenientstarting point with diameterinversely proportional to the samplinginterval. It is dem- for derivingapproximations that alsorespect reciprocity. Examples in- onstratedthat for largeranges Filon integrationis a naturalextension. It eludea small-slopeapproximation, a reciprocal phase perturbation ap- isdemonstrated that a trapezoidalFilon integrationis more accurate than proximation,and a reciprocalsmoothing approximation. [ Work support- higher-orderFilon schemes,which generate earlier spurious arrivals. In ed by ONR.] orderto yieldseismograms without spurious arrivals, the wavenumber or slownesssampling must be sufficientlydense. However, in someregions, thereflectivity function is quite regular. In theseregions, linear interpola- tion is adequateand the numberof samplingpoints is reducedby an adaptivescheme. Finally, in the slownessdomain as opposed to the wave- 2:30 numberdomain, the positionof the peaksin the reflectivityfunction is AA7. A numerical treatment of the fluid/elastic interface for the nearlyindependent of frequency.This enablesan efficientvectorization overfrequency. a• Present address: Department of OceanEngineering, potential equations.G. H. Knightly, G.-Q. Li, D. F. St Mary MIT, Cambridge,MA 02139. (Departmentof Mathematics,University of Massachusetts,Amherst, MA 01002), and D. Lee (Naval Underwater SystemsCenter, New London, CT 06320) RecentlyShang and Lee [.I. Acoust.Soc. Am. g$, 654-660(1989) ] 1:45 carriedout an investigationof the fluid/elasticinterface boundary. The AA4. Treatment of an elastic bottom in the parabolic approximation. developmenlwas done in sucha manneras to provideconsislency be- JohnS. Papadakis,Michael I. Taroudakis(Foundation for Research tweenthe standardparabolic approximation in thefluid medium and the and Technology--Hellas. Institute of Applied and Computational potentialequations for the elasticwave equations in the elasticmedium. Mathematics, P.O. Box 1527, 711 10, Iraklion. Crete, Greece), and The developmentis designedin sucha manneras to allowthe useof existingimplicit finite-difference marching schemes in theliquid medium. Betty Mayfield (Hood College,Frederick, MD 21701) Here,an enhancementof the interfacetreatment is presenled,stability of The useof impedanceboundary conditions in thesolution of the para- the schemesis discussed,and numericalcomparisons are madewith a bolicequation in underwateracoustics enables the treatmentof bottoms knownsolution having an interface.[Work supportedin part by ONR. ] S65 J.Acoust. Soc. Am. Suppl. 1, VoL 86, Fall 1989 118thMeeting: Acoustical Society of Amedca S65 2:45 sionalfrequency-domain propagation model based on the parabolicequa- AAS. Underwater acousticpropagation simulation using a 3-D sea- tion (PE) method.The sea-surfacescattering model computes the time- surface scattering model and a 2-D propagation model. Richard domainscattering acoustic field using multiple realizations of sea-surface S. Kciffer, Guy V. Norton, David B. King, Stanley A. Chin-Bing spectrato givea realisticstochastic 3-D sea-surfacescattering prediction. (Naval Oceanographicand Atmospheric Research Laboratory, The interfacingof the two modelsis achievedby obtainingthe output NumericalModeling Division, Stcnnis Space Center, MS 39529-5004), from the sea-surfacescattering model as a functionof modepropagation angleand redistributingthe scatteredfield into the propermodes in the and JorgeC. Novarini (SYNTEK Engineeringand ComputerSystems, Inc., Bay St. Louis, MS 39520) propagationmodel. This new field is then marchedforward by the PE modeluntil the next significantsea-surface scattering event occurs where A three-dimensionaltime-domain sea-surface scattering model based the interfacingmethod is again employed.Examples will be presented onthe wedge assemblage method has been incorporated into a two-dimen- usingvarious interfacing schemes. WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM D, 1:15 TO 2:35 P.M. SessionBB. Architectural AcousticsIV, PsychologicalAcoustics V, and Musical AcousticsIII: Distinguished Lecture David Lubman, Chairman HughesAircraft Company,P.O. Box 3310, Building600, M. S. F241 Fullerton,California 92631 Chairman's Introduction--l:15 Invited Paper 1:20 BBl. Binaural technology:Fundamentals and applications.Jens P. Blauert (Lehrstuhl ffir Allgemeine Elektrotechnikund Akustik, Ruhr Universitaet,D-4630 Bochum,West Germany) Binauralsound systems have been used in sciencefor quitesome time, for example,to provideauthentic reproductionand simulation of auditoryevents for sound-qualityevaluation. However, it isonly recently that an industrystarted to evolvethat successfullyapplies binaural methods to a varietyof practicaltasks, and suppliesthe necessary equipment for doing so: the binaural-technology industry. In thislecture, the fundamen- talsof binauraltechnology will be reviewedand a few currentapplications discussed, such as dummy-head stereophony,binaural mixing consoles, binaural room simulation, binaural noise diagnosis, virtual auditory environment,and telepresence.Special emphasis will be put on the possibilitiesof incorporatingbinaural technologyinto designprocesses in architecturalacoustics. 2:20-2:35 OpenDiscussion S66 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S66 WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM A, 1:15 TO 3:03 P.M. SessionCC. SpeechCommunication V: Neural Networks and Hidden Markov Models Howard Nusbaum, Chairman Departmentof Psychology,University of Chicago,5848 SouthUniversity Avenue, Chicago, Illinois 60637 ContributedPapers 1:15 schemeprimarily differs from the standardsilence/aperiodic/periodic classificationin thai the definingcharacteristics of glottal sourceand CCI. Neural networksfor estimatingartieulatory Imsitions from speech. burstfriction sounds do notdepend on periodicitydistinctions, but on the Bishnu S. Atal and Olivier Rioul (Acoustics ResearchDepartment, locusof the energyconcentrations. One malespeaker reciting the Rain- AT&T Bell Laboratories,Murray Hill, NJ 07974) bow passagehas been recorded and analyzed.Utilizing a strictly layered This talk describesan applicationof neuralnelworks for estimating backpropagationnetwork, the automatedlearning procedure is trained positionsof variousarticulators (such as the tongue,the lips,etc.) in the usingthe first 30 s of the passage;the final 75 s are usedfor testing. vocaltract fromthe speech signal. In general,a neuralnetwork consists of Quantitativeresults, along with severalexamples illustrating this model, a largenumber of interconnectedcomputational elements. The networks will bepresented. Additionally, the performance and classification strate- thatwill bediscussed in thispaper include an inputlayer of nodesconnect- gy of the networkwill be comparedto that of humanobservers. [Work eddirectly or throughan intermediatelayer of hiddennodes to an output supportedby AFOSR. ] layer.Iterative gradient search procedures are often used for determining the unknownparameters of the neuralnetwork, but theseprocedures are very slow for trainingneural networkswith a large numberof hidden nodes.For estimatingarticulator positions, it wasfound that the weights 1:51 in the first layercould be setto fixedrandom values during the training procedurewithout degrading the performanceof the network.The ran- CC4. Context-modulateddiscrimination of similar vowelsusing second- dom fixedweights in the first layer permit the useof a fast noniterative order eonnectionistnetworks. Raymond Watrous (Department of procedurefor determiningthe unknownparameters of the secondlayer. Computer Science,University of Toronto, Toronto, Ontario M5S I A4, Testson a vocaltract modelwith ten articulatoryvariables show that the Canada) articulatorpositions can bedetermined accurately, using a networkwith Discriminationof mediciadjacent vowels in the contextof voicedand 500 hiddennodes in the intermediatelayer, from ten LPC parameters derivedfrom the speechoutput of the model. unvoiecdstop consonants using connectionist networks is investigated. Separatediscrimination networks were generated for one speakerfrom samplesof thevowel centers of [e,a•] for the sixcontexts [b,d,g,p,t,k]. A singlecontext-independent network was similarly generated.The con- text-specificerror rate was 1%, whereasthe context-independenterror ! :27 was 9%. A method for merging isomorphicnetworks into a single cc2. Acousticfeature developmentduring unsupervisedlearning by a network is described.The methoduses singular value decompositionto neuralnet. BradleyS. Seebachand Nathan Intrator (Center for Neural find a minimalbasis for the sel of context-specificweight vectors. Con- Science,Box 1843,Brown University, Providence, R102912 and Division text-dependentlinear combinationsof the basisvectors may then be com- of AppliedMathematics, Brown University, Providence, RI 02912) putedusing second-order network units.Compact networks can thus be obtainedin which the voweldiscrimination surfaces are modulatedby the A biologicallyplausible neural network model that employs unsuper- phoneticcontext. In a preliminaryexperiment, as the numberof basis visedlearning was applied to varioussets of CV syllables.This network vectors was reduced from 6 to 3, the error rate increased from 1% to 3%. has beenshown to developrecognition of input signalson the basisof Experimentswith nonlinearoptimization of context-modulated,second- distinctivesignal features rather than overallsignal shape [N. lntrator order networksfor this taskare underwayand will be reported. and B. Seebach,Int. Neural Network Soc. Abstr. 1, Suppl. 1, 299 (1988) ]. Syllablespronounced in isolationby maleand femalespeakers weredigitized and sampledin short (8-32 ms) overlappingtime win- dows,then filteredinto overlappingcritical bandwidths[ E. Zwicker, J. Acoust.SOC. Am. 33, 248 ( 1961) ] to producethree-dimensional energy 2:03 surfacesin timeand frequency. A portionof thesesyllabic tokens was used asa trainingset for thenet. Those remaining were used to testgeneraliza- CC5. Automated languageacquisition. A. L. Gorin and S. E. Levinson tion of networksolutions both within a singlespeaker's utterances and (SpeechResearch Department, AT&T Bell Laboratories,Murray Hill, acrossspeakers. For example,when trainedon a singlespeaker's tokens, NJ 07974) and testedfor classificationof placeof articulationin stopconsonants, the networkmighl correctlyidentify approximately 80% of similar tokens A newapproach to developinglarge-vocabulary speech understand- from a differentspeaker. ingsystems, in whichthe system automatically acquires a language model for its task basedon semantic information, is introduced. This is in con- trast to previouswork, in which languagewas preprogrammed. An im- portantconsequence of this approach is that it leadsto habitablelanguage models.To accomplishthis task,use is madeof a medium-grainneural 1:39 network,together with a noveladaptive training procedure for estimating CC3. A eonneetionistmodel for classifyingspeech into silence,glottal the set of connectionweights. The resultingconnection weights have an source, burst frietlon, or mixed categories. Steven J. Sadoff (Central information-theoreticinterpretation, and do not requiregradient search Institute for the Deaf, 818 South Euclid, St. Louis, MO 63110) techniquesfor their estimation.A conversational-modesystem that serves as a test bed for the network is described.The applicationscenario is An algorithmfor classifyingspeech into four classes(silence, only inward-callmanagement, where a customertelephones a large organiza- glottal source,only burst friction,or mixed) is beingdeveloped. This tion and encountersan operatorwhose function is to forward the call to S67 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of America S67 oneof severaldepartments. The network'sperformance is firstexperi- Weiner filters, one for each stateof the hidden Markov model for the clean mentallyevaluated on threetext databases:sentences collected from the speech,where the weightingfactors at eachtime instanceare the proba- inward-callmanagement system, transcriptions of customerresponses to bilitiesof the filtersto be the correctfilters for the givennoisy speech. The an automatedcall transactionsystem (ACTR), anda collectionof idio- other MMSE estimators considered here have similar structure. At 10-dB maticphrases connoting yes and no. Finally, preliminary results evaluat- input signal-to-noiseratio (SNR), the proposedapproach provided an ing the inward-callmanagement system using speech data comprising improvementof 4.5-5.5 dB in SNR. unconstrainedsequences of isolatedwords, which are phonetically tran- scribedvia an ergodic HMM andprovided as noisy input to thenetwork, are reported. 2:39 CC8. Speechrecognition using HMMs with an LVQ-trained codebook. Hitoshi lwamida, Shigeru Katagiri, and Erik McDermott (ATR, 2:15 SanpeidaniInuidani, Seika-cho Soraku-gun, Kyoto 619-02,Japan) CC6. HMM-based word recognitionusing word category prediction A newspeech recognition algorithm using HMMs with a codebook neural network. Katsuteru Maruyama, Masami Nakamura, trainedby learningvector quantization (LVQ) is heredescribed. Both Takeshi Kawabata, and Kiyohiro Shikano (ATR Interpreting HMMs and LVQ arestochastic algorithms holding considerable promise TelephonyResearch Laboratories, Sanpeidani Inuidani Seika-chou, for speechrecognition. In particular,HMMs havethe significantadvan- Souraku-gun,Kyoto 619-02,Japan) tagethat phonemodels can easilybe concatenatedto producea long utterancemodel, suchas a word or sentencemodel. LVQ, on the other Word categoryprediction is usedto implementan accurateword rec- hand,is a powerfulclassifier, as shown in the highphoneme recognition ognitionsystem. Traditional statistical approaches require considerable ratesobtained in McDermott and Katagiri (1989). The new algorithm trainingdata to estimatethe probabilitiesof word sequencesand many describedhere combinesthe advantagesinherent in eachof thesetwo parametersto memorizeprobabilities. It is alsodifficult to predictunseen algorithms.To evaluatethis LVQ-HMM hybrid,phoneme recognition data.To solvethis problem,NETgram (Neural Network for word cate- experimentswere performed using the same data as used in TDNN (Wai- goryprediction) has been proposed. The performanceof the NETgram bel, 1988) and shift-tolerantLVQ (McDermott and Katagiri, 1989) ex- wascomparable to thatof thestatistical trigram model [ M. Nakamuraet periments.Applied to variousphoneme recognition tasks, the LVQ- aL, Proc. IEEE ICASSP '89, p. 731, s13.10 (1989.5) ]. In order to evalu- HMM hybrid achievedrecognition rates much higherthan thoseof a ate the abilityof the NETgram, the NETgram is appliedto an English conventionalHMM with a codebookdesigned by K-meansclustering. wordrecognition experiment based on HMM. Englishphone models for For smallcodebook sizes, the hybridmodel realized a morethan tenfold HMM weretrained using the forward-backwardalgorithm according to decreasein recognitionerror rates, yielding extremely high phoneme rec- thepronunciation symbols of trainingspeech. The NETgramis efficiently ognitionperformance, comparable to that of TDNN or shift-tolerant appliedto improveword recognitionresults using a dynamicprogram- LVQ. mingtechnique. The word recognitionrate usingHMM is 81.0%. The NETgram (trigrammodel) trainedby 512 sentencesin the BrownCor- pusText Databaseimproves the word recognitionrate to 86.3%. The statisticaltrigram modelimproves the word recognitionrate to 85.4%. The NETgram performsbetter than the statisticaltrigram modelal- thoughthe training data are insufficient to estimatethe probabilities of the trigramword category sequence. This experimentshows the NETgram 2:51 canlearn general characteristics from limited training data, that isto say, the NETgram can interpolatesparse training data to a certaindegree. CC9. A new discriminative HMM training procedure. Eng- FongHuang • and Frank K. Soong(Speech Research Department, AT&T Bell Laboratories, 600 Mountain Avenue, Murray Hill, NJ 07974) A trainingprocedure is proposedfor improvingthe discriminative 2:27 powerof a maximumlikelihood (ML) hiddenMarkov model(HMM) withoutsacrificing its classificationcapability. The proposeddiscrimina- CC•/. Minimum mean-squareerror speechenhancement using hidden tive HMM consistsof a conventionallytrained ML modeland a discrimi- Markov models.Yariv Ephraim (SpeechResearch Department, AT&T nativemodel. The trainingdata are utilized in twodifferent modes. In the Bell Laboratories,600 Mountain Avenue, Room 2C-572, Murray Hill, first mode, conventionalML models, denoted as master models, are NJ 07974) trained. In the second mode, discriminative models, denoted as slave A minimummean-square error (MMSE) estimationapproach for models,are trained by aligningtraining tokens of a certainword with all enhancingspeech signals degraded by statisticallyindependent additive but the correctword mastermodels and the modelparameters are esti- noiseis developed.This approachis motivatedby the fact that optimal matedby maximizingthe conditionallikelihood of the trainingtokens waveformcoding in the MSE sense,and optimalautoregressive model giventhe fact that they are aligned with incorrect-word master models. In vectorquantization in theItakura-Saito sense, given noisy speech, can be recognition,the scores of themaster models are reinforced by the scores of performedusing MMSE estimatorsof thespeech waveform and its sam- theinput tokens, which are compared with corresponding slave models. A plespectrum, respectively. Estimators for theclean speech signal, and for speaker-independent,39-word, alpha-digit database was used to evaluate thespectral magnitude and sample spectrum of thatsignal, are developed, the new trainingprocedure. Experimental results indicate that the new basedupon hidden Markov modelingof the cleansignal and the noise trainingcan improve the recognitionperformance by 1%-2%. However, process.The parametersof the modelsfor the two processesare estimated the discriminativepower of the slavemodels decreases gradually when from trainingsequences of cleanspeech and noisesamples. It is shown moresophisticated models and features are used. • On leavefrom Tele- that the MMSE estimatorof the cleansignal comprises a weightedsum of communicationLaboratories, Chung-li, Taiwan. S68 d.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S68 WEDNESDAY AFTERNOON, 29 NOVEMBER 1989 ST. LOUIS BALLROOM D, 3:30 TO 5:30 P.M. Plenary Session Harvey H. Hubbard, Chairman President,Acoustical Society of America BusinessMeeting Presentation of Awards DistinguishedService Citation to Daniel W. Martin Silver Medal in Bioresponseto Vibration to Floyd Dunn SilverMedal in EngineeringAcoustics to JoshuaE. Greenspon Silver Medal in Musical Acoustics to Max V. Mathews Entertainment "Bingham'sMissouri" will be presentedby the St. Louis Chapterof Young Audiences,Inc. This national award-winningpresentation, featuring music from the MissouriFolk Trio, chroniclessights and soundsalong the Missouri River during the 1800s,as viewedby celebratedAmerican artist GeorgeCaleb Bingham. S69 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of America $69 THURSDAY MORNING, 30 NOVEMBER 1989 DIRECTORS' ROW 46, 8:00 TO 11:45A.M. Session DD. Animal Bioacoustics I: Animal Bioacoustics and Noise William C. Cummings,Chairman OceanographicConsultants, 5948 Eton Court, San Diego, California 92122-3203 Invited Papers 8:00 DDI. Animalpsychoacoustics. Richard R. Fay (ParmlyHearing Institute, Loyola University of Chicago, Chicago,IL 60626) Thesense of hearingin animalshas been studied quantitatively using a varietyof behavioraltechniques for thepast 50 years.The database now existing [ R. Fay,Hearing in Vertebrates:A Psychophysics Databook (Hill- Fay Associates,Winnetka, IL, 1988)] includesdata for 180vertebrate species and allows several generaliza- tionsto be maderegarding the relationsbetween human and animalhearing, the methodologiesuseful in animalhearing studies, and the hearing processes common to all vertebrates.Although the frequency range of hearingvaries by many octaves among vertebrates, absolute sensitivity at bestfrequency is quite similar ( ñ 10 dBre: 0.0002 dyn cm 2).The sound discrimination performance often, but not always, exceeds the perfor- manceof animallisteners. There are no clear phylogenetic trends toward more acute discrimination abilities in primates,and very few examples of specializationsfor enhancedsound discrimination abilities among verte- brates(echolocators may be exceptions).There is no singleconditioning or psychophysicalmethod that clearlygives lower or morevalid thresholds. Data fromthe literature will bepresented to illustratethese and othergeneral conclusions about animal hearing. 8:25 DD2. Echolocationsignalsofodontocetesinol•enwaters.Whitlow W. L. Au (NavalOceanSystems Center, Kailua, HI 96734) The characteristicsof echolocation signals used by the Atlanticbottlenose dolphin ( Tursiopstruncalus) andby the beluga or white whale (Delphinapterus leucas) in target detection and discrimination experiments in openwaters will bediscussed. Echolocation signals measured in openwaters have been found to haveconsider- ablydifferent properties than those measured in tanks.The primaryreason for thedifferences in echolocation signalsmay be attributed to theuse of very-low-intensitysignals by animalsin tanksand the use of higher- intensity( > 40 dB) signalsin openwaters. Signals are projected in theforward direction along a narrowbeam andechoes are detected by a receivingbeam that overlaps the transmission beam. Acoustic properties of these signals,such as sourcesound-pressure levels, signal waveform, frequency spectrum, peak frequency, bandq width,and click intervals, will beconsidered. The acoustic characteristics of echolocation signals propagating fromthe animal's head will alsobe discussed. The structureof theacoustic field directly forward of theanimal, the transitionregion between the nearand far acousticfields, the directivitypatterns in the verticaland horizontalplanes, and off-axis distortion of signalswill be addressed. 8:50 DD3. Experimentalverification of a spectrogramcorrelation theory for FM echolocation.James A. Simmons (Departmentof Psychologyand Sectionof Neurobiology,Division of Biologyand Medicine,Brown University, Providence,RI 02912) The FM bat, Eptesicusfuscus,perceives each reflectingglint in a target as the cross-correlationfunction of its echoes,using a mixtureof time- and frequency-domainauditory representations according to the time separationof echocomponents relative to the receiver'sintegration time of 350,us. The accuracyof jittered- delaydiscrimination is as goodas 10 ns.Separate jitter discriminationcurves for 0ø and 180ø phase shifts of echoesfit togetherin the mannerof compoundperiod histograms to form the completecross-correlation functionof echoes.Jitter discrimination curves shift along the delay axis by - 17/•s/dBwhen echo amplitude changes(amplitude-latency trading), showing that the bat usestiming of neuraldischarges rather than extran- eous,spectrally encoded cues to discriminatejittered echoes. Other recentjitter experimentsmissed part of theseresults due to anunintended masking effect. A spectrogram-with-phaserepresentation of echoesfollowed by a transformationof the frequency-domaincomponent of the spectrograminto the time domainwould accountfor thebat's perception of the cross-correlationfunction. This transformationis the basisfor fusing targetrange and targetshape into unifiedimages. These results additionally implicate vertical and horizontal localizationas part of a singleimaging process. S70 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall1989 118thMeeting: Acoustical Society of America S70 ContributedPapers 9:15 SPL. At eachintensity-duration combination, the levelof the decrement was varied to obtain a decrement threshold. The minimal detectable de- DIM. Sonar signal design of the Chinese bat Pipistrellus abramus crementdecreased from approximately20 dB at the shortestduration to ternminkand the French bat Myotis mystacinusduring natural hunting an asymptoteof roughly4 dB at approximately30 ms. The data were prey.Zhenbiao Lin, fin Lu (Departmentof RadioEngineering, South modeledby a low-passfilter with an I l-ms time constant.The decrement China Universityof Technology,Guangzhou, People's Republic of detectionfunction of the chinchillais similar to that of humans.However, China), and BernardEscudie (Laboratory of SignalProcessing, ICPI, longduration decrement thresholds are largerin the chinchilla,as would 25 rue du Plat, Lyon, France) bepredicted from the largeintensity difference limen of thechinchilla. In Thereare more than 100species of batsin China.This papergives the general,there was little changein the decrementfunction across back- firstdescription of ChineseFM bat signalsand a comparisonwith the groundintensities except that 2-msdecrements were not detected at the25 FrenchFM bat. Twentysignals emitted by a SouthChina bat, Pipistrellus dB SPL background intensity. [Research supported by grants abramusternmink (P.a.t.), and49 signalsby a Frenchbat, Mgotis rnysta- RO 1NS 16761 and RO I NS23894. ] cinus(M.m.), duringsearching, identification, pursuit, and prey capture in the field were studiedvia time, frequencyanalysis, and ]-D wideband ambiguitydiagrams. The evolutionof mainsignal parameters (duration, dominantfrequency, bandwidth, and averagesampling pace along the huntingtrack) wasdescribed as a functionof estimatedbat-prey distance in order to find the relation betweenthe parametersand distance.The DD6. An old cochlearfrequency.position function comparedto more resultssuggest that theP.a.t. bat and the M.m. bat usea similarstrategy of recentdata. D. D. Greenwood(Audiology/Speech Sciences, University signaldesign although they are living in differentcontinents, and hunting of British Columbia, Vancouver,British Columbia V6T IW5, Canada} differentinsects in differentenvironments. For two speciesof FM bats,a turningpoint in theparameter evolution exists at about1.5 m betweenbat Accuratecochlear frequency-position functions based on physiologi- and prey.In our recordedsignals, the M.m. bat emitsmuch larger band- cal data wouldassist interpretation of physiologicaland psychoacoustic width and strongerharmonics than the P.a.t. bat. As a consequence,the datawithin and across species. In 1961,a familyof almost-exponential rangeresolution and bearingresolution are higherfor the M.m. bat than functionswas developed[D. D. Greenwood,J. Acoust.Sac. Am. 33, for the P.a.t. bat in thisexperiment. [ Work supportedby NSFC 1881155 1344-1356 (1961)] by integratingan exponentialfunction fitted to a of China and RCP 445 CNRS of France; partially supportedby NIH subsetof critical bandwidthestimates. The integralfitted well cochlear grant NS 20527 to ProfessorP. Jen of the Universityof Missouri--Co- data on frequency-placein humancadavers, and, with changesin con- lumbia, Columbia, MO. ] stants,similar data on severalother species--as well as behavioral-ana- tomical data on cats. Since 1961, mechanical data from dead human ears andvarious data from livinganimals (eat, guineapig, chinchilla, monkey, gerbil) haveappeared. It is shownthat theyare fittedby the samebasic function.As in 1961, in both cadaverand living ears, the sameslope parametercan be usedif partitionlength is normalizedto 1;i.e., slope is 9:30 scaledif distanceis expressedin physicalunits. Slope uniformity suggests that the endaverdata may well describeat least the function'sslope in DDS. Detection of intensity decrementsby the chinchilla. Christine livinghuman cochleas. To setthe upperfrequency limit requiresonly the J. Graf, SamuelS. Saunders,and Richard J. Saivi (Hearing Research Laboratory,Department of CommunicativeDisorders and Sciences, empiricalchoice of a singleother parameter. It is suggestedthat these StateUniversity of New York at Buffalo,Buffalo, N¾ 14214) cochlearfrequency-position functions be used in plottingnot only physio- logicaldata but alsopsychoacoustic data {and in modelingwork with Three manauralchinchillas were trained to detectintensity decre- speechand other bioscaustic signals), since, if the cochleardata are cur- mentsin broadbandnoise (20 kHz) usinga shock-avoidancecondition- feet, the functionsare consequentlyindependent of, and an appropriate ing procedure.The intensitydecrements were presentedat one of nine meansto examine,psyehoacoustic data and assumptions.[Work sup- different durations between2 and 35 ms at noiselevels of 25, 45, and 65 dB portedby NSERC, Canada.] 10:00-10:15 Break 10:15 earlyproportional to the frequency and the speed was approximately con- stant.The storageand lossmoduli were also approximately constant in DD7. Measurement of the law strain complex extensional modulusof this frequencyrange, with meanvalues of 179300 Pa and 113700 Pa, rabbit skin. J. Michael Pcreira, Joseph M. Mansour, and Bryan respectively.[Work supportedby NIH.] R. Davis(Department of MechanicalandAerospace Engineering, Case WesternReserve University, Cleveland, OH 44106) The complexextensional modulus of rabbitskin was measuredby recordingacoustic pulses propagating along thin stripsof skin excised from the backsof rabbits.In the experiment,a stepchange in displace- 10:30 mentwas introduced into the skinsample by a speaker,to whichone end DDS. Improved spatial resolutionof fish using hydroacoustics.P. of the samplewas attached. The responsewas recorded at differentloca- H. Patrick, B. Sim, and G. Hunt (Biological ResearchSection, Ontario tionsalong the sampleby a phonographpickup. The nominalstatic stress Hydro ResearchDivision, 800 Kipling Avenue, Toronto, Ontario on the samplesduring the experimentswas maintainedat a low level M8Z 5S4, Canada) ( 1500Pa). Repeatable•sults for theattenuation, speed, and viseoelastic storageand lossmoduli were measuredin the frequencyrange of 200- Investigationswere conducted in the laboratoryon theaccuracy of a 1000Hz. In thisfrequency range, the attenuation was approximately lin- commercialsonar system in detectingunderwater targets. Accurate spa- S71 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S71 tial resolutionof fishis necessaryfor reliableestimates of biomassand for cientvariable upon which to basenoise exposure standards. [ Work sup- classifyingfish species. Results suggested that the sounderwas limited in portedby NIOSH.] targetseparation below 25 cm at a transmittedpulse of 0.1 ms.Modifica- tionsto equipmentresulted in an improvedresolution (to approximately 10cm) of targetsin bothvertical and horizontaltank experiments.Target strengthestimates were also determined. The applicationof hydroacous- ticsto detectand classifyfish is discussed.[ Work supportedby Ontario Hydro. ] 11:15 DDll. Effects of sodiumsalicylate and intensenoise on thresholdsin the chinchilla. B. R. Bancroft, F. A. Boettcher,and R. J. Salvi (Hearing 10:45 ResearchLaboratory, SUNY--Buffalo, Buffalo, NY 14214) Salicylateadministration exacerbates the temporarythreshold shift DD9. Ineffectivenessof ultrasonic flea collars. Michael W. Dryden, Glenis R. Long, and Sayed M. Gaafar (Departmentsof Veterinary (TTS) causedby noiseexposure in humans [D. McFadden and H. S. Pathologyand Audiologyand SpeechScience, Purdue University, West Plattsmier,Hear. Res. 9, 295-316 (1983) ]. However, resultsfrom animal Lafayette,IN 47907) experimentsare conflicting as to whethersalicylate administration poten- flaresnoise-induced TTS. Becauseof the controversyin the literatureand Ultrasonicflea collarsfrom two companieswere evaluatedfor their the lack of data on the potentialpermanent interaction of noiseand salicy- ability to reduceflea numberson catswith experimentallyinduced flea lates,chinchillas were exposedto noise(105 dB SPL, 0.5-kHz octave infestations.Sounds produced by the collarswere also evaluatedboth band, 15days), salicylate( 300 mg/kg, daily i.p. injection),or both,con- before and after use to ensure that the flea collars were functional. Each currently. Salicylateadministration resulted in a TTS of 4 to 8 dB, de- brand of ultrasonic flea collar was evaluated on five cats with each cat pendingon frequency,and no permanentthreshold shift (PTS). Noise beinginfested with 50 cat fleas (Ctenocephalidesfelis)for a 7-day treat- exposurecaused 46 to 68 dB of TTS and 10 to 26 dB of PTS. The group ment and controlperiod. Cats weredeclawed and fittedwith Elizabethan exposedto bothagents received 57 to 70 dB of TTS and 8 to 30 dB of PTS. collarsto reducenormal grooming activity. Collars generated 40-kHz Therewas little differencebetween the interaction and noise-alone groups pulses(0.8 -ms exponential rise and fall ) every50 or 20 ms ( dependingon for both TTS and PTS measures.In conclusion,salicylate did not exacer- the device). The level of the stimuli was between 92 and 80 dB SPL at 10 bate the temporaryor permanentthreshold shifts caused by intensenoise cm and wouldthus be audibleto both catsand dogs.An averageof 99.6% exposure.[Work supportedby NIH 5ROI-NS23894-03.] and 97.4% of the fleas were still on the cats after treatment and control periods,respectively. The ultrasonicflea collars were totally ineffective at reducingflea numbers on thesecats. 11:30 11:00 DD12. The threshold recovery pattern between 48 h to 30 days after exposure to 50-I.tS arc-dischargeimpulse noise in guinea pigs. DDI0. Kurtosismeasurements: Implications for noiseexposure criteria. Wenyuan GaoandFanming Yuan (DepartmentofNavyMedicine,The R. P. Hamernik, K. D. Hsueh, W. A. Ahroon, R. I. Davis, and G. SecondMilitary MedicalCollege, Shanghai, People's Republic of China) A. Turrentine (Auditory ResearchLaboratory, 107 Beaumont, State Universityof New York, Plattsburgh,NY 12901) The thresholdrecovery pattern 48 h after exposureto arc-discharge impulsenoise was investigated. Guinea pigs, implanted with chronicelec- Two conventional measures of an industrial noise environment are the trodes,were measuredin quiet usinga click average-evokedresponse overallsound-pressure level (SPL) andthe power spectrum. While these technique.The animalswere exposed to ten impulses(l/min) of either two measuresare necessary,they may not be sufficientto describea noise 170-(N=10), 166-(N=17), 163-(N=15), 160-(N=9), or 157- environmentfor the purposesof evaluatingthe hazardto hearing.This (N= 16) dB peakSPL. The A durationwas 50/zs, and the risetime was presentationwill describethe results of an experimentdesigned to testthe 10/•s. Recoverythreshold was monitored,at 48 h, at 8, 15, 22, and 30 hypothesisthat for equalSPL and powerspectrum a high-kurtosisnoise days, respectively,after exposure.Some of the animalswere then sacri- exposureis more hazardousto hearingthan is a low-kurtosisnoise. Two ficedfor cochleahistology. The resultswere as follows.( 1) The recovery groupsof chinchillas,with two animalsper group,were exposed continu- patternsvaried with peakamplitude of impulsenoise, i.e., theywere linear ouslyfor 5 days,to oneof twoclasses of noiseat a 90 dB SPL.The power in log time of 157and 160dB; they werenonmonotonic in log time of 163 spectrumof bothnoises was identical but the noiseshad differentvalues of and 166dB, andthere was no recoverybut a growthpattern at 170dB. (2) kurtosis.The resultsclearly showed that therewas up to a 20 dB greater The periodbetween 8 to 15days post exposure is an importantperiod for permanenthearing loss for theanimals exposed to thehigh-kurtosis noise. turning to permanentthreshold shift (PTS); i.e., if the recoverygoes on in Detailed resultsof asymptoticand permanent thresholdshifts and sen- this period, it will show no PTS. (3) This kind of impulsenoise with short sorycell losseswill be presented.These results would appearto be a clear duration and rapid risetime may be more harmful to the audition than the indicationthat, even for moderatelevels of exposure,energy is nota suffi- other kind. S72 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America $72 THURSDAY MORNING, 30 NOVEMBER 1989 ST. LOUIS BALLROOM G, 8:00TO 11:50A.M. SessionEE. PhysicalAcoustics V: MechanicalWave Propagation in CondensedMatter Physics Philip L. Marston, Cochairman Departmentof Physics,Washington State University,Pullman, Washington99164-2814 Moises Levy, Cochairman Physics'Department, University of Wisconsin,Milwaukee, Wisconsin53201 Chairman's Introduction4:00 Invited Papers 8:05 EEI. Ultrasonicabsorption in themagnetic superconducting system Er,_• Ho• Rh4B4.KcunJ. Sun(NASA- LangleyResearch Center, Mail Stop231, Hampton,VA 23665) Ultrasonicattenuation measurements as a functionof temperatureat constantmagnetic fields and as a functionof magneticfield at constanttemperatures in Er, xH%Rh4B•show interesting behavior at low temperatures.The enhanced attenuation observed in thesuperconducting state and the sharp change in attenu- ationat magneticphase transition for the superconductors in this system arc attributed to spin-phonon interac- tion. The peculiardependence of attenuationon orientationof magneticfield with respectto soundwave propagationdirection evidences indirectly the coexistence of magneticorder and superconductivity. A two- Icyelmodel of phonon-phononinteraction is proposed to elucidatethe occurrence of relaxationtype attenu- ationmaxima observed in thetemperature-dependent attenuation curves of all thesamples. Furthermore, the attenuationbehavior at low temperaturesin high magneticfields may alsomanifest the occurrenceof a new magnetoelasticcoupling effect. 8:35 EE2. Ultrasonic attenuation measurementsou the heavy fermiou superconductorUPt 3. B. K. Sarma, M. Levy, A. Schenstrom,Y-J. Qian, M.-F. Xu (Department of Physics, University of Wisconsin- Milwaukee, Milwaukee, WI 53201), S. Ademwalla,J. B. Ketterson,Z. Zhao (Department of Physics, NorthwesternUniversity, Evanston,IL 60208), and D. Hinks (MST, Argonne National Laboratory, Argonne,IL 60439) The superconductingpairing mechanism in the heavyfermion superconductor UPh is unconventional,as evidencedfrom any of its low-temperatureproperties. Recent ultrasonic absorption, heat capacity,and mag- netizationmeasurements show a multiplicityof phasesin theH-Tplane, suggestingan exotic order parameter. The experimentalresults will be reviewed.[Work supportedby ONR and NSF. ] 9:05 EE3.Ultrasonic studies of collective modes in superfluid '•He. John B. Ketterson(DepartmentofPhysicsand Astronomy,Northwestern University, Evanston, IL 60201) Liquid3He undergoes a transition ata temperatureof2 X 10-3 K intoa superfluidstate. The state is related to thepaired-electron state of theBCS theory of superconductivitybutdiffers in thatthe '•Hc atoms arc paired in an L = I ( p wave),S = 1 (triplet) state[rather than the L = 0 (s wave),S = 0 (singlet)state of conven- tionalsuperconductivity]. Various excited states of the superfluidexist, which arc referredto as cxcitonicor collectivemodes, and, loosely, they correspond to variousvalues of thetotal angular momentum J- L + S. Someof thesemodes couple strongly with ultrasound. A briefreview is presented of thesecollective modes and of theultrasonic experiments performed to date,which probe their behavior as a functionof temperatureand magneticfield. $73 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S73 9:35 EE4. Wave localizationin anisotropicrandom media. Ping Shcng(Exxon Researchand EngineeringCo., Route 22 East, Clinton Township,Annandale, NJ 08801) As a phenomenongeneric to wavesin randommedia, localization has many general features common to both quantumparticles, e.g., electrons,and classicalwaves, e.g., electromagnetic and elasticwaves. One such aspectis that all waveslocalize in two- or one-dimensionalsystems with an arbitrary amount of randomness, and that in three dimensionsa wave localizesonly in certain energy regimesthat arc separatedfrom the alelocalizedregimes by so-calledmobility edges. In thispaper, the resultsof recentwork on the anisotropic dimensionalcrossover behavior for wavelocalization are described. Starting from a randomlylayered medium, somescattering centers were introduced,i.e., inhomogeneities,into eachof the layersin a controlledmanner. Theseinhomogeneities cause the waveto scatterout of the layeringdirection. When the strengthand the densityof thescattering inhomogeneities become equal to therandomness encountered in propagatingfrom layerto layer,then the system becomes an isotropic,3-D randommedium. Therefore, in increasingthe scatter- ing strengthand densityof the scatterersthere is essentiallya "crossover" from one-dimensionalrandomness, whereall wavesare localized,to isotropicthree-dimensional randomness, where there can be mobilityedges. What isfound is that thereis a criticalanisotropy below which the system behaves as 1 D andabove which the systembehaves as 3 D. In other words, the transition is achievedin a discontinuousmanner. The talk will emphasizethe underlyingphysics of the localizationand its anisotropiccritical transition. 10:05 EES. Surfaceacoustic waves on nonlinearsubstrates. Alexei A. Maradudin (Departmentof Physics, Universityof California, Irvine, CA 92717) The nonlinearityof the substrateon whicha surfaceacoustic wave propagates eausea acoustical rectifica- tion of the surfacewave, harmonic generation and nonlinearmixing of surfacewaves, and the acoustoelastic effect.It givesrise to weaklynonlinear surface acoustic waves, and to associatedsurface acoustic solitary waves.The spatialdispersion required for theexistence of thelatter is introduced by coatingthe substrate by a thin film with differentmaterial properties. Explicit conditions for the formationof envelopesolirons are presented,as well asfor self-focusingof thesurface acoustic waves. The nonlinearitycan also cause Love waves tobecome leaky with anamplitude-dependent damping constant. Finally, a periodiccorrugation of thesurface ofa nonlinearelastic substrate isused to introduce spatial dispersion into the system. An amplitudedependence of thestop bands for thepropagation of weaklynonlinear surface acoustic waves across the resulting periodic structure,and an amplitude-dependentattenuation in the caseof sagittalpolarization, are found. [Work supportedby NSF Grant No. DMR 88-15866.] 10:35 EE6. Relating macroscopicdeformation to site symmetrychanges around the Cr + + + in shockedruby crystals.Y. M. Gupta (Departmentof Physics,Washington State University, Pullman, WA 99164-2814) An importantbut very difficultproblem in shockwave research is to understandand interprettime- resolved,continuum measurements in condensedmaterials in termsof changesat theatomic/molecular level. Asa starttoward this challenging problem, experimental and theoretical studies on understanding the lumines- cenceR-line spectrain shockedruby crystalshave been carried out. Time-resolvedexperiments have been performedto obtainwavelength shifts under macroscopically, well-defined compression and tension loading, andfor shockwave propagation along the crystal c anda axes.Below the Hugoniotelastic limit ( • 13GPa), the observedR-line spectraare stronglyanisotropic and shownonlinear shifts with densityfor bothcompres- sion and tension.For the theoreticalanalysis, the deformationpotential is written in terms of symmetry adaptedbasis according to the irreduciblerepresentation of localoctahedral group. Using the symmetry of the macroscopicstrain, the perturbative effects of the deformation potential are evaluated for the -'Estate that gives riseto the R lines.Results not only from the shockexperiments but alsofrom hydrostaticand uniaxialstress experimentscan he analyzed consistently using these analytic developments. A summary of theexperimental andanalytical work will be presented.[This workwas carried out in collaborationwith S. M. Sharma,X. A. Shen,P. D. Horn, and J. A. Butt and supportedby ONR ContractN00014-86-0307. ] 11:05 EE7. Ballisticphonon mediated detection of elementaryparticles in silicon.B. Neuhauser(Department of Physicsand Astronomy,San Francisco State University, San Francisco, CA 94132), B. Cabrera,A. T. Lee, B.A. Young (Departmentof Physics,Stanford University, Stanford, CA 94305), C.J. Martoff (Department of Physics,Temple University, Philadelphia, PA 19122),and J.P. McVittie (Centerfor IntegratedSystems, StanfordUniversity, Stanford, CA 94305) Whenan incidentparticle collides with an electronor nucleusin a siliconcrystal, 70% of therecoil energy is convertedinto a burstof low-energyphonons. These phonons propagate ballistically for distancesof several S74 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America $74 centimetersif thecrystal has very few defects or impuritiesand if it iscooled to a temperalumwell below l K. Direct sensingof the phononsis the basisfor a newtype of elementaryparticle detector, called a siliconcrystal acousticdetector (SiCAD), whichwill be usedin neutrinoexperiments and in the searchfor otherweakly interactingparticles. The firstgeneration of ultrasensitivephonon detectors consists of a thin film stripof a superconductingmaterial, which is current biased just below the critical temperature. Incident phonons drive a segmentof the film normaland thereby produce a voltagepulse. These devices have sufficient sensitivity and resolutionto detect energy depositions of severalkeV in standard3-in. silicon wafers. Research currently is in progresson the nextgeneration of phonondetectors that will makeuse of superconductingtunnel junctions. [Work supportedby DOE, NSF, and ResearchCorporation. ] ContributedPaper 11:35 pie.Swept frequency acoustic spectra of single-crystal tantalum standing wave resonancesat the normal mode frequencies(mechanical reso- EES. Using a SQUID acoustomagneticspectrometer to characterize nances)are studiedat 4.2 K. Dependingon the rateof frequencysweep continuouswave acoustic responseobservations. Ronald K. Sunallots througha mechanicalresonance, the appliedacoustic power, and the and Mark R. Holland (Departmentof Physics,Washington University, magneticfield, the observed structure at the mechanicalresonance fre- St. Louis, MO 63130) quenciescan be bothabove and below the baseline.The low-frequency sideof eachstructure has a Lorentzline shape and the high-frequency side A SQUID acoustomagneticspectrometer is usedto detectchanges of hasan exponentialdecay that is explainedby heatexchange between sam- magnetizationof tantalummetal produced by a continuouswave compos- pie and surroundings.This interpretationallows these structures to be ite resonatorin an external magneticfield. Acoustic power variations usedin generalin continuouswave velocity of soundand ultrasonic at- producetemperature variations in thespectrum and thus cause changes in tenuationstudies while making use of thehigh sensitivity of theSQUID theCurie susceptibility of theparamagnetic impurities present in lhesam- acoustomagneticspectrometer. THURSDAY MORNING, 30 NOVEMBER 1989 ST. LOUIS BALLROOM BC, 8:00 A.M. TO 12:00 NOON SessionFF. SpeechCommunication VI: Synthesis,Perception, and Recognition(Poster Session} SigfridD. Soli, Chairman The HouseEar Institute,256 SouthLake Street.Los Angeles, California 90057 All posterswill bedisplayed from 8:00 a.m. to 12:00noon. To allowall contributorsan opportunity to seeother posters,contributors of odd-numberedpapers will beat theirposters from 8:00 to 10:00a.m. and contributors of even-numberedpapers will be at their postersfrom 10:00a.m. to 12:00noon. ContributedPapers FFI. Languagemodeling for very-large-vocabularyspeech recognition. tax, the parseridentifies many erroneous candidates and estimatesloca- D. O'Shaughnessy,V. Gupta,'} M. Lennig,'• F. Seitz, and tionsto insertpunctuation in theoutput text. The combinedparser and P. Mermelstein'}(1NRS-Tclecommunications, Nuns Island,Quebec trigrammodel have raised the accuracyof this systemto 92% for ten H3E I H6, Canada) speakerson an 86 000-word vocabulary.[ Work supportedby NSERC. ] a)Also with Bell-NorthernResearch, Nuns Island, QuebecH3E IH6, Virtually all recognitionsystems, both researchand commercial, Canada. placesevere limits on the vocabularyand syntacticstructures a speaker may use.This recognitionsystem is the most advancedin the world in termsof freedomfrom this constraint--it allows an essentiallyunrestrict- ed Englishvocabulary and sentencestructure•although it is currently requiredthat eachword be followed by a shortpause. Many speechrecog- FF2. A database demonstratingthe soundsof the world's languages. nition systemsmake little useof syntacticand semanticinformation. A Peter Ladefogedand John Choi (PhoneticsLaboratory, Linguistics trigram languagemodel is employedhere that examinesthe statistical Department,UCLA, Los Angeles,CA 90024-1543) likelihoodof eachvocabulary word's occurrence as a functionof the two precedingwords. A global context-sensitiveparser is also used which A databaseillustrating the full rangeof the soundsof the world's accepts,as input, a longsequence of wordcandidates. (The speakeris not languageswould be an enormousundertaking. Consequently, this paper requiredto pronouncepunctuation marks.) Using rulesof Englishsyn- is restrictedto tryingto provideillustrations of a largenumber of the less S75 J.Acoust. Soc. Am. Suppl. 1, VoL 86, Fall 1989 118thMeeting: Acoustical Society ofAmerica S75 familiarsounds. Currently, samples of about50 languageswere obtained potentialASR applicationsover the telephonenetwork and identifythe illustratingdifferent airstream mechanisms (!X66 clicks,Navaho ejec- sourceof the degradation.To conductthis experiment,a databasebased tives,Igbo voicelessimplosives, etc.), phonationtypes (Sindhi voiced on an 1l-digit vocabularywas created from 20 nativeEnglish speakers in aspirates,Bruu tense phonations, etc. ), andplaces and manners of articu- 11sessions in a 2-monthperiod. The utteranceswere digitally recorded on lation (Agul epiglottalstops, Yeletnye bilabial trills, etc.). Each record- two tracks simultaneously:one from a head-mountedShure SM-10A dy- inghas been digitized at 22 000Hz, with8-bit samples, care being taken to namicmicrophone, and the otherfrom oneof I 1 selectedtelephones that maintainthe full 48-dB signal-to-noise(S/N) ratio. (However,some of will be differentin eachsession. A commericalspeaker-dependent recog- the originalfield recordings do not havethis S/N ratio.) The databaseis nitionsystem was tested with boththe subdatabaseM•C created from the organizedas a 15-megset of MacintoshHypercard stacks. Clicking on a Shure SM-10A microphoneand PHONEfrom the 11 telephones.The word,or a row,or columnof words,reproduces the corresponding items, quantitativeanalysis of the evaluationtests will be presented. andmakes it simpleto repeatcontrasting sounds. The entire stack is in the publicdomain, and is available(for Macintoshcomputers with Hyper- card) at cost.[Work supportedby NSF.] FF6. A learning word-spottingmethod for speaker-independentword recognition in noisy environments. Hiroshi Kanazawa and Yoichi Takebayashi (Research and Development Center, Toshiba FF3. LPC vector quantization based on critical-band analysis. Corporation1, KomukaiToshiba-cho, Saiwai-ku, Kawasaki 210, Japan ) Daniel Lin and Brian M McCarthy (SpeechResearch Department, InternationalMobile Machines,2130 Arch Street, Philadelphia,PA A learningword-spotting method has been proposed for the purpose 19103) of robustspeaker-independent word recognitionin noisyenvironments. In orderto avoidword boundary detection errors at the recognitionstage, Critical-bandspectrum is known to representthe redundancies of the the methodemploys word spotting based on the multiplesimilarity meth- humanauditory system. Vector quantizationconstitutes a minimal re- od, whichwas shown to be effectivefor noisyspeech data. The learning dundantcode set for a givensignal source. In thisstudy, these techniques processuses synthesized noisy speech data, a mixtureof purespeech data arecombined in a low-ratespeech analysis/rcsynthesis system. An impor- and noisedata, to designreliable word referencevectors for the word tant designcriterion for a perceptuallybased source coder is to definean spotting.Word featurevectors with maximummultiple similarity values appropriatemapping between the analysiscodebook in critical-banddo- are automaticallyextracted by the word spotting.During the learning main and the synthesiscodebook in LPC domain. Severalmethods for process,the signal-to-noiseratio (SNR) of the synthesizeddata is gradu- generatingand searchingthe synthesiscodebooks based on the perceptual ally decreasedto performword spottingaccurately. Experiments were distancecriterion are proposed. Their performancewith otherLPC-bascd carried out for 13 words, including ten Japanesedigits, spokenby 50 distancemeasures (e.g., log area, LSP, cepstral) is compared.These re- males. Under the condition of 10 dB, SNR contaminatedby concourse suitsindicate a substantialimprovement in thequality of syntheticspeech noise,the recognitionscores of 85.5% and 94.1% wcrcobtained by word for the perception-basedsystem due to smootherspectral trajectories in spotting,without learningand with learning,respectively. The results the low-frequencyregions. haveshown the effectivenessof the proposedmethod in noisyenviron- ments. FF4. Evaluation of a glottal ARMA modelingscheme. Arthur P. Lobo and William A. Ainsworth (Department of Communication and FF7. Goal-directed speechin a spoken languagesystem. Alexander Neuroscience,University ofKecle, Keele,Staffordshire ST5 5BG, United I. Rudnicky(SchoolofComputcrScicncc, Carncgic-MellonUniversity, Kingdom) Pittsburgh,PA 15213) In many speechanalysis/synthesis schemes, the sourceof excitation The adventof reliablespeaker-independent continuous speech recog- for voicedspeech is a train of impulses.Although this is a mathematically nitionsystems has made it possibleto designsystems that usespeech as a soundscheme, there is no physiologicalbasis for it. The qualityof speech replacementfor keyboardinput. To understandthe natureof a system that hasbeen attained due to the introductionof a dynamicallyvarying thataccepts spontaneous goal-directed speech (as opposed to thecurrent source,e.g., a parametricsource mode, multipulse excitation, etc., has standardof readspeech), a spoken-languagespreadsheet was implemen- beenfound to be betterthan that producedusing impulse excitation. In tedand users performing a series of tasksusing this system were studied. thispaper, a pitchsynchronous glottal ARMA analysis/synthesisscheme The systemwas instrumented to allow the collectionof detailedtiming is proposed.A parametricvoice source model is discussed.The voice informationabout the componentsof the interactioncycle. The (HMM- sourceand vocaltract parametersare simultaneouslyestimated. The AR based)recognition system incorporates a lexiconof 273 wordsand a lan- and MA orders of the vocal tract model are variable and their values are guageof perplexity51. Four usersperformed a seriesof 40 tasks(involv- decided every period dependingon whether a resynthesisefficiency ing the entry of personalfinancial information) alternatingvoice and thresholdis crossed during the analysis/synthesisprocedure. This scheme keyboardinput. Userscompleted 30 tasksin one block of sessions,then iscompared with two other schemes,viz., ( 1) closedphase LPC analysis/ returneda monthlater to completethe remainder.The utterancesspoken synthesisand (2) robustLPC analysis/synthesis.The superiorityof the into the system(over 7500) were storedfor later analysis.The data col- proposedscheme over the latter two methodsis shownin termsof better lectedprovide a comprehensivepicture of howusers learn to usea spoken- formant/bandwidthtracking ability and efficiencyof resynthesis. languagesystem and providean understandingof the factorsthat govern performancein spoken-languagesystems. FFS. Transducer effect on automatic speech recognizer. Harry FF8. The intelligibilityof nativeand non-nativespeakers of American M. Chang (NYNEX, 500 WestchesterAvenue, Room 1G21, White English using spelling alphabet test materials. A. Schmidt-Nielsen Plains, NY 10604) (Naval ResearchLaboratory, Code 5532, Washington,DC 20375-5000) It is generallyrecognized that the typesof transducersused to input Male and femalespeakers who were native speakersof French and speechto automaticspeech recognizers (ASR) will affecttheir perfor- Dutch, as well as nativeAmerican Engish speakers, were recorded read- mance.However, any ASR applicationover the telephonenetwork im- ing randomizedlists of the InternationalCivil Aviation Organization pliesthe use of a largenumber.of different telephones as input transducers (ICAO) spellingalphabet words (ALFA, BRAVO, CHARLIE, etc.) to ASR. It is, therefore,very importantto quantifythe performancedeg- and the digits0 to 9. The resultingtapes were subjected to a varietyof radationdue to transducer-to-transducervariation introduced in many degradations,including Gaussian noise, interruption by silenceand by $76 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S76 noise,and LPC coding. Non-native speakers were less intelligible than the rallyoccurring variations (i.e., normal,slow, fast, soft, loud, angry). nativespeakers, and the decrement was greater for some of thedegraded Analysesshow that the acoustic characteristics of individual words vary speechconditions. Results are compared with selected DRT scoresfor the considerablyacross talkers, and across styles within talkers. Performance samespeakers. [Work supportedby NAVSPAWARSYSCOM.] of humanlisteners and the two machine-based recognition systems was testedin a single-talker,multistyle condition, and in a multitalker,multi- stylecondition. All testswere conducted under two listening conditions: normal,and in the presence of maskingnoise. The data to bepresented are theerror patterns of human listeners, versus the machine-recognition sys- FF9. Modelinglexical stress in readand spontaneous speech, Joseph tems,exhibited across talkers, across speaking styles, and across training H. Polifroniand Alexander I. Rudnicky(School of ComputerScience, conditions(multitalker, multistyle training versus single talker, single Carnegie-MellonUniversity, Pittsburgh, PA 15213) styletraining). [Work supportedby BoeingAerospace and Electronics. ] Althoughprosodic information has long been thought important for speechrecognition, few demonstrations exist of its effective use in recogni- tionsystems. Lexical stress information has been shown to improve recog- nition performanceby allowingthe differentiationof confusablewords (e.g., Rudnickyand Li, DARPA Workshopon SpeechRecogn., June FF12. Homovocalic patterns in two- and three-syllable words: 1988).In thisstudy, lexical stress modeling for a spreadsheetsystem with Implicationsfor automaticspeech recognition. Mary ReginaSmith significantnumber of confusehiewords (e.g., EIGHTY and (Bellcore,444 HoesLane, 4C-530, Piscataway, NJ 08854) EIGHTEEN) is examined. The modelsused here have beenevaluated on In 11067 two-syllablewords and 9640 three-syllablewords of the bothread and spontaneous speech. A databaseof over400 spreadsheet MRC PsycholinguisticDatabase, 5% of the formerclass and 15% of the and numericutterances was availablefor training a (HMM-based) latterhave the same vowel quality in twoconsecutive syllables. Of the 20 speaker-independentcontinuous-speech system with a 273-wordvocabu- allowedvocalic nuclei, 95 % occurhomovocalically in two-syllable words. laryand language perplexity of about51. Testing data used in thisstudy For the firsttwo syllablesof the three-syllablewords, 75% of the vowel werebased on readutterances and data generatedin a separatestudy typesoccur homovocalically, while only 45% occurin thefinal two sylla- examiningthe useof a spoken-languagespreadsheet. This latterset in- blesof three-syllablewords. Which vowels occur in thesesequences is cludes:(a) a "spontaneous"set, composed of parsableutterances from a significantlyaffected by stresspattern, as is theirrelative frequency in spreadsheettask; (b) a "read"set, consisting of the spontaneoussen- suchsequences. The likelihood that a voweltype will occur in a homovo- tencesread by theiroriginal speakers. The useof lexicalstress models was calicsequence can be expressed asa functionof quality, stress pattern, and foundto reducethe error rate for readspeech by approximately10%. A combinatoryrange. The resultingpatterns reduce the searchspace and comparisonwith the spontaneousdata will providean insightinto the promiseimproved probabilities in someautomatic speech recognition natureof the improvement. strategies.[ Work supportedby Alvey (U.K.). ] FF10. Automaticevaluation of Englishspoken by Japanesestudents. FF13. Time and frequency spectral derivative features for robust Jared Bernstein, Mitch Weintraub, Mike Cohen, and Hy Murveit recognition of Lombard and noisy speech. Brian A. Hanson, Ted (SpeechResearch Program, SRI International,Menlo Park, CA 94025) H. Applebaum, and Gregory R. De Haan (Speech Technology Laboratory, 3888 State Street, Santa Barbara, CA 93105) Thepaper describes the methods and results of a studyof thefeasibil- ity of automaticallygrading the performanceof Japanesestudents when readingEnglish aloud. SRI recorded31 adultJapanese speakers: 22 men environments.This paper addresses automatic, speaker-independent rec- and 9 women.Each Japanesespeaker read six sentencesaloud. All 186 ognitionof Lombardand noisyspeech. Spectral information was derived recordedutterances were presented in a randomorder for ratingby three fromproduction-based (LP) or perceptuallybased LP analysis,and rep- expertlisteners who rated the utterances on twooccasions. Speech-grad- resentedby cepstralor index-weightedcepstral (RPS, spectralfrequency ing softwarewas developedfrom an adaptivehidden-Markov-model derivative)coefficients. Instantaneous, dynamic (first temporalderiva- (HMM) speech-recognitionsystem. The gradingprocedure is a two-step tive),Lombard and accelerationspeechapproximates(second temporal speechvariationsderivative) encounteredfeatures were innoisythen process:First, the speech to begraded is aligned, then the segments of the computedfrom one of the representations,and integrated during the like- speechsignal that are locatedare comparedwith modelsof thoseseg- lihoodcalculation of a hiddenMarkov model recognizer. The recognizer mentsthat havebeen developed from a databaseof speechfrom native wastrained with normal speech to evaluate its robustness to speech varia- speakersof English.Important points in the resultsare: ( l ) ratingsof tions.Strong interaction was found between the temporalderivative fea- speechquality by expertlisteners are extremely reliable, and (2) automat- tures,the spectralderivative, and the degreeof smoothingin the analysis. ic gradesfrom the systemcorrelate well ( > 0.8) with thoseratings. Althoughthe acceleration feature performed poorly by itself,when com- binedwith otherfeatures it generallyraised recognition rates for Lombard speech.This trend was more pronounced with thecepstral representation. Combiningall threefeatures gave the bestLombard speech recognition results:The perceptuallybased analysis with RPS coefficientsand three featuresyielded 94% correctrecognition, compared with 63% and 82% FFI1. Wordrecognition by humansand machines: Tests on a multitalker, correctfor standardcepstrum and RPS, respectively,when the instanta- multistyledatabase. Patricia K. Kuhl, Kerry P. Green(Department of neousfeature was used alone. Experiments were alsodone with additive Speechand HearingSciences, University of Washington,Seattle, WA noise.The resultsand implicationsfor robustspeech recognition are dis- 98195),John W. Gordon,David L. Sanford,and Caroline Fu (Boeing cussed. Aerospaceand Electronics,Seattle, WA 98124) Experimentscomparing isolated word recognition by humanlistnets with automaticspeech recognition systems are valuablebecause error analysesmay leadto improvementsin speechrecognition technology. FFI4. Speaker-independentrecognition of the DARPA Naval Resource Isolatedword recognition in adulthuman listeners has been compared Management Task. Chin-Hui Lee, Lawrence R. Rabiner, Jay withrecognition performance by twocommercially available speech-rec- G. Wilpon,and RobertoPieraccini '• (SpeechResearch Department, ognition systems.The test stimuli were drawn from the Lincoln Laborato- AT&T Bell Laboratories,600 Mountain Avenue, Murray Hill, NJ ry Stressed-Speechdatabase. The databaseconsists of 6930 stimuli (two 07974) iterationsof each of 35words spoken by nine different people in 11differ- entspeaking styles). The vocabulary contains confusab]e words (i.e., go, In thistalk, theissue of acousticmodeling of thefundamental units for hello,oh, no, and zero); the speaking styles include a widerange of natu- recognitionis examined.Modeling approaches for bothcontext indepen- S77 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S77 dent (CI) and contextdependent (CD) units are studied.The acoustic FFI7. On the use of speech recognition in articulation training. modelingapproaches were tested on speakerindependent recognition of Marl Ostendorf(Boston University, Electrical, Computer, and Systems the DARPA Naval ResourceManagement Task. The setof contextinde- EngineeringDepartment, 44 CummingtonStreet, Boston,MA 02215), pendentunits in thisstudy is a fixedset of 47 phonelikeunits (PLUs), in Kerin Gibbons(Raytheon Company, 131 Spring Street, Lexington, MA which eachPLU is associatedwith a linguisticallydefined phoneme sym- 02173), Elizabeth Gavett, and Nicholas Bankson(Boston University, bol. EachCI/PLU ismodeled using a continuousdensity hidden Markov Boston, MA 02215) model(CDHMM) with Gaussianmixture state observation density. The set of contextdependent units includesPLUs definedby left, right, and Computerprocessing of speechhas been used in manyways as an aid both left and right context. Only thoseCD/PLUs with enoughoccur- to cliniciansin varioustypes of speechtherapy. In thiswork, use of speech rencesin the trainingdata are selectedfor modeling.Two approachesare recognitionfor detectingarticulation errors, such as a lispor a phoneme presentedto model the CD/PLUs. Both CI/PLUs and CD/PLUs are deletion,is investigated.The main featuresof this approachare ( 1) the obtainedusing the segmentalk-means training procedure. In the caseof useof a time-dependentscore in templatematching and (2) the adjust- contextindependent acoustic modeling, the maximumnumber of mix- mentof dynamictime-warping slope constraints according to the typeof tures was varied in each state from 1 to 256 and it was found that the word articulation error (deletionsversus substitutions). The use of a time-de- accuracy increasedfrom 56% to 89% which indicates that sufficient pendentscore is particularlyimportant because it allowsthe clinicJanto acousticresolution is essentialfor improvedperformance. The 89% word targeta particularportion of a word for correctpronunciation. The algo- accuracyis the highestperformance reported based on contextindepen- rithmwas implemented in nearreal time on a PC with a DSP processor. dent units.When contextdependency modeling is incorporated,both Speaker-dependentand -independentperformance was investigated for modelingapproaches achieved better than 92% word accuracy.a•On simulatedand actual articulation errors. [Work supportedby DesignAu- leavefrom CSELT, Torino, Italy. tomation,Inc., throughan SBIR grantfor the NationalInstitute for Deaf- nessand Other CommunicationsDisorders. ] FF15. Automated training for speech recognition. L. A. Smith, B. FFIS. Perception of synthetic vowels: A comparison of several L. Scott,L. S. Lin, and J. M. Newell (Scott InstrumentsCorp., 1111 classificationschemes. John W. Hawks (Central Institute for the Deaf, Willow SpringsDrive, Denton, TX 76205) 818 South Euclid Avenue, St. Louis, MO 63110) A methodis describedfor automatedtraining of a speaker-indepen- Syntheticvowel tokens (1725) were randomly presentedtwice to dentisolated word rccognizer. The trainingprocess generates vocabulary eightnaive speakers of MidwesternAmerican English for classificationas templatesfrom a databaseof collectedtraining utterances. These tem- one of 12 vowel categories,/IY,IH,EH,EY,AE,AA,AH,AO,OW,UH, platesare then modified through adaptive training, an iterativeprocess of UW,ER/. Subjectsrated the certaintyof their responseson a scalefrom testingand modifying templates in orderto optimizerecognition. Robust- one (very unsure)to five (very sure). The vowelswere synthesized in null nessof the templatesis enhancedby varyingthe presentationof the col- contextand utilizeda male F0 contour.The frequencyvalues off 1, F2, lectedutterances during adaptation; varying the utterancesampling rate, and F3 wereselected such that the entirearea where vowds may be repre- for example,has the effectof presentingthe sameutterance at differing sentedin the auditory-perccputalspace [J. D. Miller, J. Acoust.Soc. Am. pitchesand time scales.Adaptive training continues until the error rate 85, 2114-2134 (1989) ] was equidistantlysampled. The resultsof this fallsto an acceptablelevel. Results will bcpresented for similarvocabular- experimentsupport the viewthat F 1and F2 arethe primary determinants iesdeveloped with and withoutadaptation and undervarying adaptation in the perceptionof nonretrofiexvowels. In addition, monophthongal conditions. versionsof the diphthongs/EY/and/OW/may be distinctly classified. The resultswill be presentedgraphically as targetzones constructed on the basisof the pluralityidentifications for eachtoken. These target zones areabutting and nonoverlapping, and correctly classify 99.9% of the plu- FF16. Transitional cues and redundancyin speechrecognition. Brian rality judgments(less 48 tokenswhere ties occurred)and 75% of the L. Scott, Lloyd A. Smith, Lisan S. Lin, and J. Mark Newell (Scott 27 600 total judgments.Other vowel classificationschemes will also be InstrumentsCorporation, 1111 Willow SpringsDrive, Denton, TX comparedfor their accuracyin classifyingthe resultsof this experiment, 76205 ) as well asresults from otherstudies. [Work supportedby NINCD.] Perhapsthe mostdifficult features for speechrecognition systems to properlycapture and align are the transitionalcues that existbetween phones.There are advantagesand disadvantagesassociated with both linear time normalizationmethods and dynamictime warping (DTW). FF19. Attentional allocation during phoneme monitoring: An Linearmethods preserve the relativedurations of transitionaland steady- investigationinto the unit of perceptualanalysis and selectiveattention stateportions of the signalbut tendto smearfeatures and loseresolution. during speech perception. Mark A. Pitt and Arthur G. Samuel DTW retainsresolution but losesdurational relations. The presentalgo- (Departmentof Psychology,Box 11-A Yale Station,Yale University, rithm attemptsto compensatefor the smearingassociated with linear New Haven, CT 06520) normalizationby copyingtransitional regions of the utteranceto addi- tionallocations within the representation. This process involves locating A variantof the phonemcmonitoring technique (in whichthe proba- theonset and offset of thesyllable nucleus based on the normalizedrepre- bility with whichthe targetphoneme occurred at a particularconsonant sentationand copying two frames immediately surrounding the onset and locationin a wordwas varied) was cmployc.d to investigatewhether the offsetto anotherlocation in the representation.The processenhances phonemeor the syllableis the smallestlinguistic unit of perceptualanaly- performanceof the recognizerin two ways.First of all, the redundancy sis,and to examineselective attention during speech perception. The re- servesto weightmore heavilythe perceptuallyimportant transitional re- suits suggestthat the phoncmgmay bca unit of perception and that sub- gionthat exists between consonantal cues, such as burst frequencies, frica- jects were able to attcndselectively to a singleconsonant location in a tion, nasalmurmurs, etc., and the vowel nucleusof the syllable.And word. IResearchsupported by AFOSR.] second,copying these regions to anotherlocation within the representa- tion servesto enhancethe alignmentof thesecues during trainingand matching.The systemis beingevaluated over standardtelephone lines usingthe alphabet as the test vocabulary. The telephoneenvironment, and FF20. On the combineduse of "full" and "reduced"diphones in speech the inherentconfusibility of the alphabetas a testvocabulary, places sig- synthesis.Rob Drullman and Ren• Collier (Institute for Perception nificantdemands on the recognizerto accuratelycapture and identify Research, P.O. Box 513, 5600 MB Eindhoven, The Netherlands) transitionalcues. The pilot study,using only a singlespeaker, provided encouragingresults. The recognizerwas 96.7% accurateover 8320 live In the generationof syntheticspeech by meansof diphones,the latter trials. Currently,an alphabetdatabase is beingcollected and speaker- are usually excisedfrom accentedsyllables. Therefore, all the speech independenttest results will be presented. soundsthat result from the concatenationof diphoneshave phonetic S78 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America $78 propertiesthat are typicalof thisaccented context only. In particular, Successfulintegration of segmentaland accentual phages of phonologicai theirtemporal and spectral characteristics are close to thetarget values processingis shownto bepossible using the conceptof recurfiveadjunc- that onefinds in the productionof carefullyspoken, isolated, accented tion of a suffixto a stem.The ruleswere installedin the computerin a syllables.Speech generated by meansof suchdiphones often sounds over- human-editableformat, translated by a rule compilerinto executableLisp articulated.In unaccentedsyllables (and perhapselsewhere, too), one functions,tested, and foundto bevalid in actualrealization of synthesized mightprefer to use"reduced" diphones, which may lead to shorterand speech. spectrallyless extreme speech sounds. Perception experiments were run in orderto assessthe improvementthat canbe achievedby alsousing "re- duced"diphones, which are extracted from unaccentedsyllables. In these tests,the useof "reduced"diphones was determinedby the following variables:word stress, sentence accent, word and phrase boundaries, met- riealstructure. [ Researchsponsored by theDutch government as part of FF24. On the unit set design for speechsynthesis by rule using the nationalSPIN program"Speech analysis and synthesis."] nonuniformunits. Yoshinori Sagisaka (ATR InterpretingTelephon• ResearchLaboratories, Kyoto 619-02,Japan) The unit set designalgorithm is proposedfor nonuniformsynthesis units usingentropy measure. This algorithmenables optimal unit set buildingaccording to statisticphonotactie characteristics without using FF21. Including 'frequency-dependent losses in a time-domain, apriorilinguisticknowledge. By consideringeach phoneme sequence unit articulatory synthesizer.R. $. McGowan (Haskins Laboratories,270 asa statein nonuniformorder Markov process,the entropy of a unit setis Crown Street, New Haven, CT 06511 ) definedby the entropyof a Markov informationsource. Each unit candi- datecorresponds to a newstate and the newstate generation corresponds The Websterhorn equationdescribes acoustic wave propagation in to the enrollment of the candidate as a unit member. Unit enrollment tubeswith cylindricalsymmetry. This differentialequation can be derived takesplace by findingthe correspondingnew state that givesthe greatest from the linearizedequations of motionfor air neglectingviscosity and decreaseto the entropyof the set. Applyingthis algorithmto Japanese heat conduction.A modifiedhorn equationincluding viscous loss in the usingphoneme sequence statistics derived from a Japaneseword dierio- boundary-layerapproximation can be derived.This modifiedequation nary,the following characteristics were observed. ( I ) Linguisticallywell- includesa convolutionintegral over time; thusthe computationalprob- known CV-type syllablesand diphthongsare selectedin the very early lemis to findan approximation to theintegral. First, the integral is discre- stage.(2) Though the frequentlyused dyadic VC-type units are not se- tized,and thena rationalPad6 approximation is usedto approximatethe lectedin theearly stage, the CVCV-type unitsthat containthose VC-type resultinginfinite sum. The frequencydependence of thisapproximation is unitsare selected in theearlier stages. (3) Many earlyselected units coin- examinedand comparedto the idealfrequency dependence. [Work sup- cide with the phonemesequence contained in morphcrees.Through this portedby NIH grantsHD-01994 and NS-13617to HaskinsLaborato- experiment,it hasturned out that the nonuniformunit setcovered pho- ries.] nemesequences about two timesmore efficientlythan the usualuniform unit set. FF22. Excitation waveform extraction for pitch control in residual- excited LPC speechsynthesis. K. Itoh and H. Sato (Speech and FF25. RobustARMA analysisfor the determinationof voicesource and AcousticsLaboratory, NTT Human InterfaceLaboratories, Musashino- vocal tract control parametersin speechsynthesis. Johan de Veth, shi,Tokyo 180,Japan) Louis Bores (Department of Language and Speech, Nijmegen University,P.O. Box 9103, 6500 HD Nijmegen,The Netherlands),and The residual-excitedLPC is one of the mosteffective techniques for Wim van Golstein Brouwers (PTT Research Neher Laboratory, producinghigh-quality synthetic speech. However, this technique has dif- Leidschendam, The Netherlands) ficultycontrolling pitch frequency when applied to realizingsynthesized speechwith a pitch contourdifferent from originalspeech or whenap- A cascadesix pole-pairand •ve zero-pairsynthesizer has been devel- plied to synthesisof arbitraryspeech created with concatenatedspoken opedas part of a text-to-speechconversion system for Dutch. Control units.In previousmethods, pitch-period LPC residualwaveforms that information for this synthesizeris derived from a.o. measurementson extractedpitch synchronously have becn used to controlpitch frequency. naturalspeech. A pitchsynchronous robust ARMA analysistechnique However,the extraction position (Ep) andwindow length (E•, } arevery wasdeveloped and applied to utterancesproduced by a numberof adult criticalto synthesizedspeech quality, and causevoice quality deteriora- maletalkers. The resultingpole-zero parameters were separated into sets tion.This paper proposes a new method of excitationwaveform extraction pertainingto the sourceof the vocaltract. The vocaltract parameters thatautomatically determines E•, and E,o using spectral envelope distor- were correctedin thoseframes where the analysismethod made occa- tion criteriabetween input and syntheticspeech. Subjective evaluation sionalmistakes. Analysis-resynthesis of sentence material using the cor- experimentsindicate that the pitch frequency pattern can be changed with rectedvocal tract parametersto controlthe synthesizerdriven by impulse a relativelysmall deterioration in quality.Application of this methodto and noiseexcitation yielded high-quality synthetic speech. The tract pa- arbitraryspeech synthesis will alsobe discussed. rameterswere then used to inversefilter the speech,to obtainthe source function, that was subsequentlyparametrized using the Liljencrants- Fant model.It is hopedthat the speechquality will beimproved by replac- ing the impulseexcitation by the controllablesource model. FF23.Phonoiogicai rulesfor a semanti•s.to..speech systemofJapanese: A rule applicationexperiment. Shigeru $ato (InformariesLaboratory, Tohoku Instituteof Technology,Yagiyama-kasumicho 35-1, Taihaku, Sendal982, Japan) FF26. Restorationof tracheocsophageaivoice with LPC resynthesis. YingyongQi (Departmentof Speechand Hearing Science, University of In orderto confirmvalidity of thephonological rules of Japaneseand Arizona, Tueaon, AZ 85721 ) their applicationalgorithms, a phoneticform generationexperiment is performedin the phonologicalcomponent of a semantics-to-speechsys- Four vowels,[il, [•e], [a], and [u], and one diphthong[ou], pro- tem implementedin thecomputer. The rulesare tieredaccording, first, to ducedby two maleand two femaletracheoesophageal speakers, were ana- the natureof data they handle:segmental versus accentual, and, second, lyzedwith the LPC autocorrelationmethod. The vowelswere synthesized to their rule-applicationalrelevance to syntax:cyclic versusnoncyclic. by replacingthe originalsource with an impulsetrain. The fundamental $79 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of Amedca S79 frequencyof theimpulse train was 100 Hz for themale and 200 Hz for the tion.These results are contrasted to otherrecent studies, and methodolog- femalespeakers. The resultsof anidentification experiment indicated that ical and theoreticalaspects of parameterestimation in the AI modelare boththe voweland the genderof the speakercan be better identified from discussed. the synthesizedvowels than from the originalones. The possibilityof improvingthe qualityof tracheoesophagealspeech and building prosthe- tic devicewith the LPC techniquewill be discussed. FF30. An algorithmfor distinguishingbetween voiced stops and voiced fricatives.LaDeana F. Weigelt,Steven J. Sadoff,and JamesD. Miller (Central Institute for the Deaf, 818 South Euclid Avenue, St. Louis, MO FF27. Effects of spectralsmearing on speechperception. Mariken ter 63110) Keurs, Joost M. Festen, and Reinier Plomp (Department of An algorithmhas previously been reported which distinguished voice- Otolaryngology,Free University Hospital, P.O. Box 7057, 1007 MB Alhsterdam,The Netherlands) lessstops from voicelessfricatives with a successrate of 96.8% [J. Acoust. Soc.Am. Suppl. 1 85, S56 (1989)]. With only slightmodifications, this Connectedspeech presented in quiet is highly redundant.For sucha algorithmalso makes the voicedstop/fricative distinction. Here, results condition,the resolutionof spectralcontrasts by the ear seemsto be much are presentedon the modifiedalgorithm. The input signalis high-pass largerthan would be required. This suggeststhat theear's high selectivity filtered(cutoff frequency of 125 Hz) and two measuresof the resulting in frequencyis particularlyimportant to understandingspeech in the waveformare used: the rms energy and the derivativeof rmsenergy over presenceof interferingsounds. The effectof reducedfrequency resolution time (termedrate of rise,ROR). The ROR is usedas the primaryclassi- on the speech-receptionthreshold (SRT } for sentencesin noisewas inves- fier while energypulse duration and relativelevel are usedto discard tigatedfor eightnormal-hearing subjects by simulatingan auditorysys- spurious,irrelevant peaks. Peaks in ROR areconsidered in orderof mag- temwith variablefrequency selectivity. Signal processing was performed nitudefor relevanceto the stop/fricativedistinction. The resultingalgo- by short-timefast Fourier transforms(FFT), reductionof contrastin the rithm wastested on 420 CVC tokens(three malespeakers, three female spectralenvelope without affecting the harmonicstructure, and overlap- speakers,three stops and an affricate [B,D,G,JH], four fricatives pingadditions to reconstructa continuous signal. The Spectral envelope in [Z, ZH,V,DH], and ten vowel contexts [IY,1H,EH,AE,AA,AH,AO, the frequencyregion from 1{30to 8000Hz wassmeared over fixed relative UH,UW,ER] ) recorded in an anechoicchamber. Data from two male bandwidths of 1/8, 1/4, 1/2, 1, 2, 4, and 8 octaves.Results show that the andtwo femalespeakers (280 tokens)were used as a trainingset, and the SRT increasesprogressively as the spectralenvelope is smearedover remainingdata ( 140 tokens) were usedas a test set.The overall success bandwidthsexceeding the ear'scritical bandwidth. In a secondexperi- rate was97.9%. [Work supportedby AFOSR.] mentphoneme, confusions as a resultof threedifferent degrees of spectral smearingare studiedin nonsenseCVC syllables. FF31. The role of speechrate in social evaluation.Cynthia L. Crown (Departmentof Psychology,Xavier University,Cincinnati, OH 45207) FF28.Perception ofvowels by budgerigars (Melopsittacus undulatus). Robert J. Dooling, SusanD. Brown, Amy Nespor (Department of Prior studiesconcerned with the relationof speechrate to the person- Psychology,University of Maryland, CollegePark, MD 20742), and ality variableof extraversionand to socialevaluation (i.e., interpersonal John W. Hawks (Central Institute for the Deaf, St. Louis, MO 63110) perception)have primarily examinedmonologues. The studyreported herewas designed to (a) exploresuch dependencies in dialogues,and (b) Budgerigars(parakeets) were trained usingoperant conditioning investigatethe possibilitythat speechrate may be influencedby differ- techniquesto respondto differencesbetween speech stimuli. Response encesin interpersonalattraction. The 38 femalepairs who participatedin latencieswere used to constructsimilarity matrices and multidimensional thestudy were selected on thebasis of a sociogramdesigned to assesstheir scalingprocedures were then usedto produceperceptual (spatial) maps attractionto eachother prior to their participationin the study.Their of thesestimuli. For natural vowel tokens,budgerigars showed evidence sociogramscores divided them into threegroups: those who liked each of perceivingphonetic categories in spiteof variationin talkersincluding other, thosewho dislikedeach other, and thosewho wereunacquainted talker genderand talker age.Experiments with syntheticvowel tokens with eachother. Each pair engagedin a 30-minconversation from which generallyconfirmed and extendedthese findings. Multiple regression the speechrate of eachparticipant was electronicallydetermined. The techniquesrevealed that the perceptualdimensions obtained from MDS resultof multipleregression analyses indicated that thosewomen who were highlycorrelated with the frequenciesof the first and secondfor- spokemore quickly perceived themselves more positively, whereas those mants.These results suggest that both naturaland syntheticvowels are who spokemore slowly viewed their partnersmore positively. However, probablyperceived in similar ways by budgerigarsand humans.These variationsin rate wererelated to neitherinterpersonal attraction nor ex- resultshave relevance for theoriesof speechperception and languagede- traversion. velopmentin humans.[Work supportedby N1H.] FF32. Two paradigmsfor examiningthe role of phonologicalstress in FF29. Articulation index importance functions for contextual speech sentence processing. David W. Gow, Jr. and Peter C. Gordon materials. Theodore S. Bell, Donald D. Dirks, and Timothy Trine (Department of Psychology,Harvard University, Cambridge, MA (UCLA School of Medicine, Head & Neck Surgery, Rehabilitation 02138) Building,Room 31-24,Los Angeles,CA 90024-1794) The role of phonologicalstress in sentenceprocessing was studied The relativeimportance of one-thirdoctave frequency bands toward usinga syllablemonitoring task, as well as a new short-termmemory the intelligibility of speech in various contexts was examined. Thirty-five probe task. In both tasks, target syllables were embedded in bysyllabic young normal hearing adults heard sentencesin which the final word noun/verb homophoneswith syntacticcategory-dependent stress pat- (targetstimulus ) waseither predictable ( probability-high;PH) or unpre- terns.The stressand positionwithin a word of targetsyllables were ma- dictable (probability-low;PL) from the context of the sentence.Sen- nipulated.Syntactic context was also manipulated to examinethe role of tenceswere presented at S/N's from -- 8 to + 14dB in a noiseshaped to syntacticconstraints on theanticipation of stressedsyllables. The syllable conformto thepeak spectrum of the speech,and crossover frequencies as monitoringtask reproducedthe familiar facilitationof stressedsyllable related to the articulation index (AI) were determined by successively detection,and showedan effectof syntacticconstraint. Stress facilitation high-passand low-passfiltering the stimulito bisectintelligibility. Results wasnot foundwhen targets were word-final. These results were interpret- indicatedonly slight differencein 1/3 oct importancefunctions as the ed in termsof stress-facilitatedlexical segmentation and accessprocesses. result of context (PH vs PL), although the crossoverfrequency was sig- The short-termmemory probetask introduced in this researchwas shown nificantlydifferent and showedno interactionattributable to the noise. to be sensitiveto stress,but not to other sentential factors that were exam- Theeffect of contextin thesespeech materials was related to differencesin ined.The applicationof thisparadigm to issuesin metricalphonology and perceptualdynamic range more sothan the frequencyimportance func- short-termmemory is discussed.[Work supportedby AFOSR.] S80 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall1989 118thMeeting: Acoustical Society of America $80 THURSDAY MORNING, 30 NOVEMBER 1989 ST. LOUIS BALLROOM A, 8:30 TO 11:45 A.M. SessionGG. PsychologicalAcoustics VI, PhysiologicalAcoustics II, and SpeechCommunication VII: Sensory Aids for the Hearing Impaired Margaret W. Skinner,Chairman Otolaryngology,McMillan Hospital9901, WashingtonUniversity, St. Louis,Missouri 63110 ContributedPapers 8:•0 Sevenprelinguistically profoundly deaf childrenbetween 7 and 11 yearsof agewere trained in the perceptionof vowelduration and place, GG1. Results in children using the 22-electrodecochlear implant. P. the fricative/s/, and manner of articulation (/m/vs fo/, and/s/vs/t/) W. Dawson,P. I. Blarney,G. M. Clark, P. A. Busby,L. C. Rowland, distinctionsin the tactile-plus-hearing(TA} andtactile (T) conditions. S. J. Dettman, A.M. Brown, R. C. Dowell, F. W. Rickards, and J. Speechfeature recognition tests were conducted in theTA, T, andhearing I. Alcantara(Human Communication Research Centre, Department of (A) conditionsduring the following:pre-training, during-training, and Otolaryngology,University of Melbourne,Parkville, Victoria 3052 and post-trainingphases. Test scores in theTA andT conditionswere signifi- The CochlearImplant Clinic, The Royal VictorianEye and Ear Hospital, cantlygreater than A conditionscores for all but onetest, suggesting that 32 Gisborne Street, East Melbourne, Victoria 3052, Australia) the additionof the tactilesupplement improved the perceptionof these Fivechildren (aged 6.0 to 14.8years) out of a groupof nineimplanted features.During- and post-trainingphase test scoresin the TA and T with the 22-electrodecochlear implant (Cooblear Pry. Ltd.) have conditionswere significantly greater than pre-trainingscores suggesting achievedsubstantial open-set speech recognition scores for monosyllabic that the traininggiven was responsiblefor the improvementin feature word (30% to 72% for phoneroes)and sentence (26% to 74% for words perception.During- and post-trainingscores were similar suggesting that in sentences)tests using hearing without lipreading.Four of thesefive perceptualskills acquired during training were retained after training had childrenwere implanted during preadolesccnce and the fifth, who hada been removed. progressiveloss, was implanted during adolescence. The remainingchil- drenwho did not demonstrateopen-set recognition were implanted dur- ingadolescence after along duration of profounddeafness. Post-operative performanceon closed-setspeech perception tests was better than pre- 9:15 operativeperformance for the groupof fivechildren with open-setrecog- nition.Improvements in speechand language assessments were also noted GG4. A pitchperception study for a eo•hlearimplant user with residual for these children. The results are discussed with reference to variables hearing in the contralateral ear. R. van Hoesel and Y. C. Tong thatmay contribute to successfulimplant use, such as age of onsetof loss, (Departmentof Otolaryngology,University of Melbourne,Parkville, durationof profoundloss, age at implantation,aetiology, educational pro- Victoria 3052, Australia) gram,and the typeof trainingprovided. In order to combineelectrical and acousticsignals for multichannel cochlearimplant users with residualhearing in the unimplantedear, it is importantto determinethe degreeof overlapin pitchsensations on the 8:45 two sides.A pitchsealing experiment was performed maintaining con- stantand identical repetition rates on boththe implanted(left) and nor- GG2. Psychophysicaistudies on cochlearimplant patientsdeafened prior mal (right) car.The acousticsignals consisted of bandpass-filteredpulses to 4 years of age. P. A. Busby, Y. C. Tong, G. M. Clark, and I. {with centerfrequencies ranging from 350-850 Hz). Thesewere trans- I. Alcantara(Department of Otolaryngology,University of Melbourne, mired in synchronywith electricalpulses to oneof five selectedbipolar Parkville, Victoria 3052, Australia) electrodepairs spanning 15.75 mm of the scalatympani (from approxi- mately2.25 to 18 mm beyondthe base).The spacingbetween the two Psychophysicalstudies were conductedon eight prelinguallydeaf electrodesof eachbipolar pair was1.5 min. The experimentwas repeated cochlearimplant patients, between 5 and 23 yearsof ageat the time of for repetitionrates of 112.5and 205 Hz. Pitchsensations on the left (elec- surgery,who were deafened prior to 4 yearsof age.Electrical stimulation trical) sidewere reported to beas low as,and at timeseven lower than, on of the auditorynerve was by meansof the Nucleusmultiple-electrode theright (acoustic} side. All signalswere sealed with respectto anacous- prosthesis.The averagedifference limens for the prelingualpatients in tic referencestimulus that had a bandpassfilter centerfrequency at 450 electriccurrent level (7% of the dynamic rangebetween threshold and Hz. Syntheticvowels were derivedfrom the pitch sealingresults and comfort level), duration ( 17 ms for a standard of 30 ms), and the detec- acousticstrategies only as well ascombined electrical and acoustic strate- tionof a gapin a 1-s signal (9 ms) werein thesame range as those reported gieswere tested. It wasfound that the bestresults were obtained from a for postlinguallydeafened adult implantpatients. The averagedifference combinedelectrical and acoustic strategy where the electrode information limensin repetitionrate for theprclingual patients ( ! 9% for a standardof ischosen such that pitch perception overlap with theacoustic side is mini- 100 rep/s and 28% for a standardof 200 rep/s) werepoorer than those mized. [Work supportedby NIH Grant NS21027.] found for the postlingualpatients (5% for 100 rep/s and 8% for 200 rep/s). Individualdifferences in performancewere observed for the pre- lingualpatients in all studies. 9:30 GGS. Loudnesssummation study on multielectrodepair stimulation.H. H. Lira, Y. C Tong, R. D. Hollow, A. E. Vandull,and R. V. Hoesel GG3. Speechfeature recognitionby profoundlydeaf childrenusing the (Department of Otolaryngology,University of Melbourne, Parkville, University of Melbourne multichannel electrotactile aid. Joseph Victoria 3052, Australia) I. Alcantara, L. A. Whitford, P. J. Blarney,R. S.C. Cowan, and G. M. Clark (Departmentof Otolaryngology,University of Melbourne, This psychophysicalstudy investigated loudness summation for the The Royal Victorian Eye and Ear Hospital, 32 GisborneStreet, East activationof two, three,or four electrodepairs in a stimulusperiod. The Melbourne, Victoria 3002, Australia) electrodepairs were activatedin temporallynonoverlapping manner. : $81 J. Acoust.$oc. Am. Suppl. 1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America $81 Two multichannelcochlear implant patientsparticipated in the study. Sciences, Australian National University, Canberra 2601, ACT, Themethod of magnitudeestimation was used. Loudness growth for each Australia), G. M. Clark, and R. van Hocscl (Department of electrodepair wasfirst determined. The loudnessesof eachelectrode pair Otolaryngology,University of Melbourne,Parkville, Victoria 3052, in a multielectrodepair stimulus were always set equally loud. The results Australia) showedthat: (a) loudnessincreased with current level and the number of electrodepairs within a stimulus;(b) loudnesssummation for stimuli Two of the mostimportant electric signal parameters for presenting containingtwo electrodepairs was completewhen the electrodepairs usefulacoustic speech information to the Universityof Melbourne/Nu- wereapproximately four electrodesapart (3.00 mm); and (c) loudness cleusmultichannel cochlear implant are electrodeposition and electric summationfor stimulicontaining three or fourelectrode pairs was incom- pulserate. The possibilityof usingpulse rate to encodeacoustic speech pletein comparisonto the resultsobtained for stimulicontaining two informationwas investigated in thisstudy. A setof 16 (4 X 4) orthogonal electrodepairs. [Work supportedby NIH ContractNOI-NS-9-2400 and stimulicomprising all possiblecombinations of fourdifferent pulse rates NIH Grant NS21027. ] on eachof two fixedbipolar electrode pairs was presented to threecoch- lear implantrecipients for identification.Tests for perceptualindepen- dencein terms of informationtransmission indicate that the two pulse 9:45 ratespresented on eachof the two electrodepairs were perceived as par- tiallyseparable dimensions. It maytherefore be possible to useindividual GG6. Perceptual independenceof pulse rates used in speechcoding pulserates on differentelectrode pairs in a speechcoding scheme to pres- schemesfor a multichannelcoehlear implant. W. K. Lai, Y. C. Tong ent furtheracoustic speech information to a multichanneleochlear im- (Departmentof Otolaryngology,University of Melbourne,Parkville, plant. [Work supportedby NIH Contract NOI-NS-9-2400 and NIH Victoria 3052, Australia), J. B. Millar (ResearchSchool of Physical Grant NS21027. ] 10:00-10:15 Break 10:15 tion thannon-open-set listeners to recognizephoneroes and words. More importantly,a predictability-by-listenerinteraction demonstrated that GG?. Test materialsfor evaluatingspeech perception ability of cochlear open-setlisteners were better able to make useof contextualcues than implant patients. ¾umiko Fukuda (Research Institute, National werenon-open-set listeners. Rehabilitation Center for the Disabled, Namiki 4, Tokorozawa 359, Japan) A systemfor assemblingthe testmaterials that are used in evaluating speechperception ability by combineduse of cochlearimplant and lip- 10:45 readinghas been developed. In orderto constructbalanced lists, a data- baseof Japanesebasic vocabulary furnished with specificationsregarding GG9. The effect of varying the amplitude-frequencyresponse on the theirauditory, visual, temporal, and linguisticcomponents is utilizedin maskedspeech-reception threshold for sentencesin conditionswith low- thissystem. Factors whose distribution are balancedbetween each other frequency noise. Janctte van Dijkhuizen, Joost M. Festen, and list are the kind of vowel and the manner/placeof articulationof the Reinier Plomp (Department of Otolaryngology, Free University consonant,special Japanese syllables (elongated vowel, choked sound, Hospital,P.O. Box 7057, 1007MB Amsterdam,Thc Netherlands) and syllabicnasal), phones having the samemouth shape, number of syllablesin a wordand phrases in a sentence,class of familiarity,and Withina studyon themerits of a frequency-dependentautomatic gain grammaticaldata such as part of speech.All thesedata and specification controlin hearingaids, the beneficialeffect of adaptingthe amplitude- of categoryof meaningof wordsin thedatabase are used in searching frequencyresponse to situationsof seriouslyinterfering low-frequency suitablewords for the lists,as well as in analyzingcharacteristics of per- noiseis studied. For 12normal-hearing and 12hearing-impaired listeners, ceptionability of thepatients. Problems in requirementsfor theevalua- themasked speech-reception threshold (SRT) for sentencesis measured. tion methods,construction of balancedlists from the database,and video The noisehas a spectrumidentical to the long-termaverage spectrum of recordingsof speechand mouth shape of utterancesof thelists are dis- the sentences,but with its levelin oneoctave band raised by a constant20 cussedin this paper. dB in thefirst experiment, and slowly increasing by 20 dB duringsentence presentationin the secondexperiment. Results indicate that, in conditions with steady-statenoise, selective attenuation of the signalin the band containingthe extra noise gives an improvement of speechintelligibility in termsof speech-to.noiseratio of up to about4 dB for both the normal- hearingand hearing-impairedgroup. In conditionswith time-varying 10:30 noise,a similarimprovement was observed for bothgroups. GGS. A gatingtask to measurethe contributionof contextualinformation in speechperception by cooblearimplant wearers. Teresa A. Zwolanand Thomas D. Cartell (Children's Memorial Hospital, Department of Communicative Disorders, Chicago, IL 60614 and Northwestern University, Department of CommunicationSciences and Disorders, 11:00 Evanston, IL 60202) GGI0. Simplified tones for profoundly deaf tone-languagespeakers. Contextualcue utilization by two groupsof cochlearimplant wearers Teresa Ching (Department of English, ChineseUniversity of Hong wasmeasured with a gatingtask. Group I consistedof implantwearers Kong, Shatin,N.T., Hong Kong) whowere able to perceiveopen-set speech (unpracticed, unfamiliar words and sentences),while group 2 consistedof implant wearerswho were The larynx frequencypattern of speechcontributes significantly to unableto perceiveopan-set speech. Oated stimuli consisted of high-and speechcommunication when it is dependenton lipreading,especially in a low-predictabilitysentences taken from the speechperception in noise tone-languagecontext. Tests using syllable-based pitch contrastsas well (SPIN) test.Subjects' responses were evaluated to determinethe amount as connectedspeech in controlled interactivesituations show that recep- of acousticinformation required for phonemeand word recognitionas tive skillscan be considerablyimproved both in Mandarin and Cantonese wellas the percentageof phoneroesand words identified in eachtype of (two Chinesedialects). On the basisof findingswhich show that a sinu- context. Resultsindicated that open-setlisteners required less informs- soidalpresentation of voicepitch changesis morebeneficial to a pro- S82 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall1989 118thMeeting: Acoustical Society of Amedca S82 roundlydeaf person than the complete acoustic signal IS. Rosenand A. J. Talker. Subsequentconsonant confusion studies with two hearing-im- Fourcin,Speech, Hearing and Language, U.C.L. No. 1 (1983) ], thepres- pairedadults, one using tactual input in combinationwith lipreadingand ent work adoptsthis simplificationapproach in an attemptto enhance onein combinationwith aidedresidual hearing and lipreading,confirmed prusodicskills in profoundlydeaf Cantonesechildren. Results indicate that addedtactual information reduced the proportionof mannerconfu- that productiveand perceptualskills can be significantlyenhanced by sionsand to a lesserextent voicing confusions. usinga simplifiedauditory presentation. These have important implica- tionsfor the designof signalprocessing aids and the developmentof speechskills in theprofoundly deaf tone-language speaker. 11:30 11:15 GGI2. Evaluation of a 32-channel electrotactile vocoder. GGII. Speech feature encoding through an electrotactile speech Kathleen Rakowski, Christine Brenner, and Janet M. Weisenberger processor.Robert S.C. Cowan,Peter J. Blarney,Joseph I. Alcantara, (Central Institute for the Deaf, 818 S. Euclid, St. Louis, MO 63110) LesleyA. Whirford,Graeme Clark (Departmentof Otolaryngology, Previousevaluations of laboratoryprototype multichannel tactile aids University of Melbourne, Parkville, Victoria 3052, Australia), and haveshown them to provideuseful speech information. However, few of Geoff Plant (National Acoustics Laboratories, Sydney, NSW, Australia) theseprototypes have been successfully converted into wearable devices. The presentstudy evaluated a devicethat hasrecently become commer- Efficiencyof speechfeature encoding by the Universityof Melbourne ciallyavailable, the Audiotact (Sevrain-Tech, Inc.), a 32-channelelectro- multichannelelectrotactile speech processor was assessed through perfor- tactilevocoder worn on the abdomen.Four subjeerspacticipated in the manceof fight subjectson a dosed-settactual test battery. The battery study,two of whomhad previousexperience with othertactile aids. In consistedof 12 subtests,each assessingdiscrimination of a particular addition,a visualdisplay designed as a trainingaid for newusers of the speechfeature in an ABX paradigm.All subjectsscored significantly Audiotactwas evaluated. This display provides a visualrepresentation of abovechance on suprasegmentalfeatures such as syllable number, stress, the activityon the tactilearray. Two of the subjectswere instructed to use andvowel length. Discrimination of vowelformant contrasts was signifi- this visualinformation while wearingthe Audiotact.Results of testing cantly abovechance for sevensubjects. Discrimination of consonantman- with minimalpairs and smallset phonemeidentification tasks showed ner variedacross specific subtest contrasts, with betterencoding of the that mannerfeatures were transmitted quite well. Subjects were also able high-frequencyfricatives and affricates.Discrimination of initial voicing to integrateinformation available through lipreading with information contrastssuggested that tactual encodingof this featurecould be im- from the Audiotact.However, preliminary results for connecteddis- proved.Comparison of theseresults with test batteryscores from five coursetracking with the Audiotact showedonly a small improvement differentsubjects with fivecommercially available tactile devices showed overlipreading alone. The results of thisstudy are discussed in terms of superiorencoding of vowelformant and manner contrasts by the Tickle comparabilityto othermultichannel tactile devices. THURSDAY MORNING, 30 NOVEMBER 1989 DIRECTORS' ROW 43, 8:30A.M. TO 12:05P.M. Session HH. Structural Acoustics and Vibration IV: Structural Vibration and Acoustic Radiation Jerry H. Ginsberg,Chairman Schoolof MechanicalEngineering, Georgia Institute of Technology,Atlanta, Georgia 30332 Chairman's Introduction---8:30 ContributedPapers 8:35 analyzea caseof forcedvibration of a submergedshell. The resultsare Hill. Axisymmetric vihration of spheroidalshells for hlgh-aspectratio. comparedwith Yenand Dimaggio's finite difference solution [I. Acoust. Sec.Am. 41, 618-626 ( 1967}]. [Work supportedby theOffice of Naval Pei-Tai Chen and Jerry H. Ginsberg (School of Mechanical Research,Code 1132-SM.] Engineering,Georgia Institute of Technology,Atlanta, GA 30332) Assumedmode expansions based on Hamilton'sprinciple were em- ployedin an earlier presentation[P. T. Chen and J. H. Ginsberg,I. Acoust. Sec. Am. Suppl. I 85, S139 (1989)] to study fluid--structure 8:50 interactionsera submergedprelate spheroidal shell of low-aspectratio. In HH2. Responseof an infinitely long cylindrical shell with a conycoted that approach,a straightforwardanalysis of the in oacuoshell vibration internal fluid. O. Leyrat and J. M. Cuschieri(Department of Ocean propertieswas found to be efficientfor the lowerand upperbranches of Engineering,Florida Atlantic University, Boca Raton, FL 33431) axisymmetricdeformation, provided that the aspectratio of the spheroid is not large.The presentpaper addresses methods by whichthe first few The responseof an infinitelylong cylindrical shell excited by a point modesof bothbranches may be determined for casesof high-aspectratio. force, with an internal heavy medium (water) moving with a constant In order to overcomenumerical ill-conditioning, an iterative variational flow speedis examined.The solutionis obtainedby linearizingthe differ- methodis employed to seepthe natural frequencies and mode shapes with entialequations describing the behaviorof the fluid and the structureby increasingmode number. In addition,in orderto ensureconvergence to usinga spatialFourier transformin the axial directionand a modalde- theupper branch, which occurs in a muchhigher frequency band that that compositionin the circumferentialdirection. The inverseFourier trans- of the lower branch,an artificial potentialenergy affecting only high form of the solutionof the linearset of equationsgives the coupledsolu- wavenumbers is added. The model data derived in this manner are used to tion for the shell/fluid system.Using this solution,input and transfer S83 J. Acoust.Sec. Am. Suppl. 1, VoL86, Fall 1989 118thMeeting: Acoustical Society of America S83 mobilityfunctions are determined for a copperpipe of radius0.025 m and A typicalsituation in structuralacoustics and vibrations is thecase of wall thickness1.5 min. For the input mobilityfunctions, the nearfield a mechanicalsource of vibrationsattached to a large,thin structure(radi- effectscombine with the corioliseffects of the movingfluid to produce ator). In general,the forceinput into the structure, responsible for unde- instabilitiesof the circumferentialmodes at velocitiesthat dependupon sired vibrations of, and radiation from, the structure, is the resultant of the modenumber. For the transfermobility functions, broadband peaks boththe outputimpedance of the sourceand the inputimpedance of the appearcaused by the phaselag between waves of differentcircumferential structure.A theoreticalanalysis of theproblem is presentedin thecase of modenumbers. These peaks are shiftedand modifiedas the flow speed a lumpedmechanical source composed of masses,springs, and dampers, changes. with one singleattachment point to a baffled,thin, rectangular,plane structure.The couplingassumes only pure transverseforce transmission into the structure.A quadrupoleapproach for the sourceassembly and a variationalapproach for the motionof the structureallow, respectively, 9.-05 the sourceoutput impedance and the structureinput impedance to be calculatedin a varietyof configurations.The forceinput into the struc- HH3. Responseof a fluid-filled sphericalshell submergedin an infinite ture, kinetic energy of the structure, and radiation of sound from the fluid medium to a transient acoustic wave. Thomas L. Geers and structurecan then be derived.It is shownhow the natural frequenciesof Peizhen Zhang (Department of MechanicalEngineering, University of the combinedsystem can be found from the natural frequenciesof the Colorado, Boulder, CO 80309) sourceand structuredecoupled. Numerical results are presentedfor the Transientresponse problems involving fluid-filled shell structures particularapplication of optimalsuspension design of mass-typesource. submergedin infinitefluid mediaarise in variousfields, including medi- [ Dr. I. Michaluk is acknowledgedfor fruitful discussions.] cine,defense, and materials engineering. A canonicalproblem of thisclass possessesa spherical geometry, for which no solutionsapparently exist. This paperdelineates a relativelysimple formulation and method of solu- tion for thiscanonical problem and presents solutions for excitationby an 9:50 incidentstep wave. The formulationbegins with the familiarequations of motionfor a thinspherical shell and the wave equation for the internal and tIH6. Activecontrol of flexuralpower flow in _beamsusing piezoceramic externalfluid domains.The Laplacetransform is invoked,and the usual actuators.Gary Gibbs and Chris Fuller (Department of Mechanical separationof variablesmethod yields modal expressionsinvolving Le- Engineering, Virginia Polytechnic Institute and State University, gendrepolynomials and modified spherical Bessel functions. The explicit Blacksburg,VA 24061) expressionsfor the latterare thenmanipulated to yield,after transform inversion,delayed differential equations in timefor each response mode of Previousresearch has demonstrated that the flexuralpower flow in the shell-fluid system,which are integratednumerically in time. Com- thin beamscan be attenuated using active point force inputs [ C. R. Fuller pleteresponse solutions then follow by modalsuperposition, with special and L. O. Gonidou,"Active vibrationcontrol of flexuralpower flow in techniquesbeing employed to improvemodal convergence. To validate beams,"J. Acoust.Soc. Am. Suppl.I 84, S47 (1988) ]. In this paper,the themethodology, numerical results are first presented for an incidentstep resultsof experimentsthat demonstratethe feasibilityof usingsurface- waveexciting a shellwith massdensity and acousticvelocity identical to bondedpiezoceramic actuators and sensorsin controllingvibrational the correspondingproperties characterizing the sameinternal and exter- powerflow in elasticstructures are presented.The controlsystem utilizes nal fluid.Results are then shown for a step-wave-excitedsteel shell con- botha piezoceramiccontrol actuator and sensor, and a timedomain least- tainingwater and submergedin water. [Work supportedby DNA.] mean-squaresadaptive algorithm to minimizetransmitted power flow. Bothsemi-infinite and finite beams of variousgeometries are considered, andthe resultsshow significant reduction of vibrationalpower flow using only a singlepiezoceramic transducer as the control actuator. [ Work 9:20 supportedby NASA LangleyResearch Center. ] HH4. Active controlof panel-radiatednoise using multiple piezoelectric actuators. B. T. Wang, E. K. Dimitriadis, and C. R. Fuller (Departmentof MechanicalEngineering, Virginia Polytechnic Institute 10:05 and StateUniversity, Blacksburg, VA 24061) HH?. Vibrationand stabilityof an elasticbeam subjected to a periodle The potentialfor activecontrol of structurallyradiated riosic by two- axial load with time-dependentdisplacement excitation at both ends. dimensionalpiezoelectric patches bonded to the surfaceof thin elastic Xiao-FengWu and Adnan Akay (Department of Mechanical panelsis analyticallyinvestigated. A previouslypresented approximate Engineering,Wayne State University, Detroit, MI 48202) modelfor theloads induced by piezoelementsbonded to the platesurface is employedfor the calculationof the vibrationand soundradiation from This paper considersthe vibrationand stabilityof an elasticbeam simply supportedrectangular panels. The piczoelectrlcaction modifies underthe excitationof a combinedperiodic axial load and time-depen- the vibrationand henceeffects the noiseradiated by the panelwhen excit- dentdisplacement at itsends. Such problems are often encountered in the ed by constantdistributed external forces. The complexamplitude of the predictionof noiseradiation from high-speedplanar mechanismswith voltage applied to the piezoelectricis estimatedvia an optimization elasticcouplers. The governingequation of motionfor thevibrating beam schemethat minimizesthe total radiatedacoustic power into the farfield. derivedby the Hamilton's principleis highly nonlinear.To obtain an Resultshave been obtained for a singleas well as for multipleindepen- approximatesolution, the methodof assumedmodes is used.The equa- dentlycontrolled actuators. It is seenthat quitesignificant control is pos- tion of motionthen reducesto a Hill type of equationin time. The para- metric studiesof the stabilityof the beamis carried out. The stableand siblein somecases with a singleappropriately taylored actuator. The effectiveness,however, deteriorates as control of higher vibration modes unstableregions of the beam vibrationsunder the time-dependentdis- is attempted.Much betterresults are achieved with two or moreindepen- placementexcitations are investigated.The independentrestraint mo- dently controlledactuators bonded to the plate. It is clearly shownthat tionsat the endsof the beamare described by the displacementinfluence multiple,appropriately positioned actuators drastically reduce undesired functions.The total responseis finally obtainedby addingthe relative motion of the beamunder the givenboundary conditions to the specified controlspillover to the residualmodes. [Work supportedby ONR.] displacementat its ends. 9:35 10:20 HHS. Coupling effects betweena lumped vibrational source of finite impedanceand a plane radiator. Dominique Trentin, A. Berry, and HHS. Nonlinear effectsin a drivenvibrating wire. RogerJ. Hansonand Frederic Laytile (GAUS, Mechanical Engineering, University of Hilllard K. Macomber (Physics Department, University of Northern Sherbrooke,Sherbrooke, Quebec J 1K 2Ri, Canada) Iowa, Cedar Fails, IA 50614-0150) S84 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S84 A horizontalwire is driven sinusoidally with a verticalmagnetic force. 11:05 Horizontaland vertical optical detectors follow the motionof a pointon thewire. With a moderatelylarge driving force, the usual predicted non- HHII. A new efficient Ritz vectorstechnique to solve elastoacoastic lineareffects are observed as the driving frequency is increasedfrom be- problems.J.P. Coyette (Dynamic EngineeringNV, Researchand low resonance.These include the onset of a horizontalcomponent of mo- Development,Ambachtenlaan 21, B-3030Hererice, Belgium) tion, phaseshifted 90 ø with respectto the verticalcomponent, and Solutionof interiorelastoacoustic problems in the frequencydomain ultimatelyresulting in circularmotion of the wire. However,with a much is addressedthrough an efficientnonmodal approach. The procedurere- weakerdriving force, the phase relation between the horizontal and verti- lies on a symmetrictwo-field formulation (displacement variables are cal componentsis no longerfixed at 90ø and becomesvery frequency selectedfor the structure,while velocity potential is usedfor the acoustic sensitivein a certainrange. It appearsto vary from 180ø, producing in- medium}. The related second-ordersystem is transformedinto a first- clined straight-linemotion, to 90ø, to 0Q, again producingstraight-line order one by doublingthe size of the system.An updatedRitz vectors motionbut with a positiveslope. Further increase of thedriving frequency processallows the generationof two coupledbases, which enablesthe causesthe verticalexcursion of the straight-linemotion to increasewhile accuratedescription of both structuraland acousticresponses. Details the horizontalexcursion changes little. A secondharmonic in the hori- about the generationof thesevectors are provided.The processrequires zontalcomponent becomes pronounced and leadsto figure-eightmotion. only the solutionof the staticproblems and the performanceof basic Theseeffects appear not to havebeen predicted and will be illustratedon orthogonalizationand scalingoperations. The resultinghigh-vectoriza- videotape. tion abilityis stressed.Comparisons with conventionalmodal synthesis proceduresare providedin orderto showthe efficiencyof the proposed approach.It is shownthat the selectionof this techniqueleads to an optimalsolution scheme. Various numerical results related to Guy's prob- lem (cavitybacked plate} are reported. Extensions to dealwith structural and/or acousticdamped cases are alsoreported. 10:35 HH9. Analysisof soundtransmission through flexible panel/frame walls. Donald B. Bliss (Department of Mechanical Engineering, Duke University,Durham, NC 27706) 11:20 Low-frequencysound transmissionthrough flexible panel/frame HHI2. Separation of a one-dimensionalbending wave field into walls is an important noisecontrol problem, particularly for propeller propagatingand standing parts based on the reflection coefficient aircraftinterior noise. The coupledstructural/acoustic problem for spa- estimation. Jukka Linjama and Tapio Lahti (Technical Research tially periodicstructures is solved using a multiple-scaleperturbation ex- Centreof Finland,VTT/INS, SF-02150Espoo, Finland) pansion,based on the scaleseparation between panel size and acoustic wavelength,and an averagingprocedure applied over the rapidlyvarying The structuralintensity methods offer a wayof determiningthe abso- (panel) scale.A compositesolution is formedfrom a smoothedglobal lute flow of mechanicalpower anywhere on the surfacesof somebasic solutionand a periodiclocal solution that accountsfor the detailedpanel structures. In one-dimensional easessuch as beams, these methods can be motions.The panelsubsystems appear in an averagedsense in the global extendedto the determinationof mechanicalimpedance and reflection solutionthrough transfer functions, obtained from the localsolution for coefficient.In thispaper, a separationof thebending vibration of thebeam distributedframe loadings and acoustic boundary conditions. Walls com- issuggested, which divides the fieldquantities into perfectly propagating posedof identicalpanel subsystems, each containing several different pan- and standingparts, based on the estimationof the complexreflection els,attached to a flexibleframe, can be analyzed. This arrangementallows coefficient.Such a divisionis hoped to openan alternativeview in quanti- for noisereduction by alternateresonance tuning (ART), in which adja- fyingthe flow and distributionof powerin beams.The techniqueis re- centpanels are tuned to resonatealternately above and below the frequen- strictedto farfieldregions of the beam.It is alsodirectly applicable in cy to be attenuated,with the resultingout-of-phase motion producing otherone-dimensional (vibrational, acoustical, etc. } fields.Properties of acousticcancellation. Solutions are presented for a modelproblem involv- the suggestedseparation are illustratedwith an experiment,and they are ing soundtransmission through a panel/framewall and comparedwith comparedto an alternativeapproach that is based on the impedance.The the exact solutionto assessthe accuracyof the perturbationmethod. relationshipof the separatedparts to the activeand reactiveintensity is Noisereduction by ART'is demonstrated by using two panels with differ- discussed.It is shownthat eachof the fieldquantities involved in power ent resonantfrequencies in eachsubsystem. [ Work supportedby NASA transmissioncan be separated,and the propagatingparts correspond to LangleyResearch Center. ] the net (active} intensity.The standingparts show properties that maybe usablein modal analysissense. 11:35 10:50 HH13. Effect of backlashand sleevebearing clearanceon the vibration HHI0. Acoustic radiation of stiflened shells covered with a layer of spectrumof high-speedspur gear boxes. Fawzi M. A. EI-Saeidyand slightly compressiblematerial. B. Lanlagnet and J. L. Guyader T. Youshimura (Mechanical EngineeringDepartment, Tokushima (LaboratoireVibrations-Acoustique INSA deLyon, BAt. 303-20, avenue University,Minamijosanjima 2-1, Tokushima770, Japan) Albert Einstein, 69621 Villeurbanne Cedex, France) An analyticalmodel is presentedto simulatethe effectsof tooth back- This work deals with the calculation of sound radiation from a shell lashand sleevebearing clearance on the vibrationspectra of high-speed coveredwith a layer of slightlycompressible material and immersedin spur gear boxes.Elasticity in bendingand torsionof the shafts,time- water.In thiscase, the assumptionof locallyreacting material is unrealis- varyingmesh stiffness, tooth backlash,and HertzJanmesh damping, fric- tic, and a more elaborate model must be used to describethe behavior of tion betweenengaged teeth, and interactionbetween gear box casing and thecoating layer. This has been achieved by solving the equations of elasti- internalsare accountedfor. During contactbetween journal and bearing, cityin cylindricalcoordinates, and by use of asymptoticexpansions. The where time-varyingcontact stiffness with considerationof dampingat layer displacementsand stressesare then coupledto the shelldisplace- contactpoint is accountedfor, bearingclearance, system vibration, and mentsand fluid loadingon the boundarysurfaces. The calculationof shell bearingforces are calculated.The analysisis appliedto a singlestage gear radiationis madeusing the normal modesof the in oacuobasic shell, and box and equationsof motionare numericallyintegrated to obtainsystem the asymptoticexpansion truncated to the secondorder. Numerical re- responsein the time domain.The responseis transformedinto frequency sultsare presented showing the influence of layerthickness, damping, and domainusing a fastFourier transform(FFT) algorithmand samplesof Poissonratio. [Work supportedby D.R.E.T. and C.E.R.D.A.N., Mini- the resultsare shownfor differentvalues of backlashand bearingclear- st•re de la Drfense.] ance. S85 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S85 11:50 stiffness(stiffness is calculatedbased on the strainenergies of the tooth, which is treatedas a cantileverbeam of involuteshape), and Hertzian HH14. On the vibrationspectrum of high-speedspur gear boxes. Fawzi meshdamping. The beatingloads are calculatedwith considerationof M. A. E1-Sacidy(Mechanical Engineering Department, Tokushima shaft oscillationsand speed,bearing radial clearance,ball centrifugal University,Minamijosanjima 2- I, Tokushima770, Japan) force,and bearingouter ring oscillations.To demonstratethe model,a An analyticalmodel is presented to simulatethe effect of toothback- singlestage gear box is analyzed, and equations of motionare numerically lashand ball-beating radial clearance on the vibrationspectrum of high- integratedto obtainsystem response in thetime domain. This response is speedspur gear boxes. Included in thismodel are elasticity in bending and transformedinto the frequencydomain using a fastFourier transform torsionof the shafts,friction between meshing teeth, interaction between (FFrr) algorithmand samples of theresults are shown for different values gearbox casing and internals, and time-varying tooth backlash, mesh of backlashand bearingclearance. THURSDAY MORNING, 30 NOVEMBER 1989 ST. LOUIS BALLROOM F, 9:00 TO 11:50 A.M. SessionII. EngineeringAcoustics IV, PsychologicalAcoustics VII, and PhysiologicalAcoustics Ill: Hugh S. Knowles Memorial Session Mahlon D. Burkhard, Chairman 1016 Raven Lane, Palatine, Illinois 60067 Chairman's Introduction--9:00 Invited Papers 9:15 II1. Designevolution of miniatureelectroacoustic transducers. David P. EgoIf,Elmer V. Carlson,Peter L. Madaffari,and August F. Mostardo(Industrial Research Products, Inc., Elk GroveVillage, IL 60007) Thispresentation traces the evolution of thedesign of miniatureelectroacoustic transducers during the last 40 years.Of particularemphasis are thosecontributions to the research,design, and development of such transducersmade by HughKnowles and his colleagues. Pioneering efforts in theminiaturization of receivers (i.e.,small loudspeakers orearphones) and microphones made possible the transition of hearing aids from the bulkybody-worn aids of thepast to thetiny all-in-the-canal aids of today.Included in thispresentation isan overviewof hisinnovations in piezoelectric, condenser (including electrct) and electromagnetic transduction technology.Mr. Knowlos'last contribution was the research and development of the miniature transducer--a hearing-aidreceiver--incorporating a special electronic circuit to reducecurrent consumption and cxtcnd battery life of a hearingaid. 9:40 II2. Trends in hearingaids: From a technicalto an audiologicaldevice. Poul Erik Lyregaard (Offcon ResearchUnit, Eriksholm,243 Kongevejen,3070 Snckkersten,Denmark) In essence,a hearingaid is an acousticalamplifier, and has,therefore, traditionally been described in technicalterms, as definedin the appropriatestandards (IEC, ANSI). The adventof prescriptivefitting methods,plus a betterunderstanding ofthe in situ acoustics, has had a significantimpact on hearing aid design, but alsoon the way in whichthe acousticalproperties are presented. Hearing aid dispenserstend to havean audiologicalrather than acoustical background, and the data to beemployed in individualhearing aid fitting arealso audiologically based. Hence, there is a strongcase for manufacturersto adopt an audiological charac- terizationof theirhearing aids, even encompassing that hearing aid adjustments be calibrated in hearingloss and UCL units.Conversions between technical and audiological data for a hearingaid will be discussedand illustrated.In part,such conversions are contingent upon hearing aid measurements performed on the KE- MAR manikin,one of HughKnowles' many contributions in thefield of hearingaids. S86 J.Acoust. Soc. Am. Suppl. 1,Vol. 86, Fall 1989 118thMeeting: Acoustical Society ofAmerica S86 10:05 II3. AddressingHugh Knowles'fundamental question: What would"high fidelity"mean for a hearingaid wearer?Mead C. Killion (EtymoticResearch, 61 Martin Lane,Elk GroveVillage, IL 60007) Firstassigned by Hugh Knowlesto updateHugh's earlier estimate of thedifference between real-ear and 2-era• coupler frequency-response curves for hearing aids, the author later did his Ph.D. research on the topic of thedesign and evaluation of high-fidelityhearing aids. Throughout this research, Hugh's probing questions wereboth usefuland sometimesembarrassing. He madeit impossibleto givesuperficial answers to difficult questionsabout ( 1) the occlusion-effect{"hollow voice") problem with closed-moldfittings, (2} whatgain andfrequency response would be required for a perceptionof high-fidelitysound by someone with a substantial high-frequencyhearing loss, and (3) whatdirectional/frequency-response accuracy in the hearingaid would be requiredfor a "high fidelity"sense of the auditoryspace. Partly, as a resultof Hugh'squestions and contributions,a good solution has now been obtained to theocclusion-effect problem (reinventing Zwislocki's 1953solution that he thoughtprobably impractical}, there are now transducersand "acousticplumbing" techniquespermitting 40-Hz to 16-kHz reproductionin hearingaids with fidelitycomparable to that of professionalstudio monitors, in-the-canal hearing aids with directionalproperties comparable to that of the unaidedear, and a new integrated-circuithearing-aid amplifier chip with level-dependenthigh-frequency emphasisthat is designedto providenearly "transparent" amplification to the wearer.Recorded demonstra- tionswill be playedif time permits. 10:30 I14. Electroacoustictest methodsfor evaluationof hearing-aidperformance. L. B. Beck (Veterans Administration,Medical Center, 50 Irving Street,N.W., Washington,DC 20422), E. D. Burnett(National Institutefor Standardsand Technology, Gaithersburg, MD 20899),and G. D. Causey(Catholic University of America, Washington,DC 20064) Thepurpose of thispaper is to describe the development ofelectroacoustic test procedures for evaluation of hearingaids by the Veterans Administration (VA). Thesemethods have evolved from the use of the2-cm '• couplerwith puretones as the input stimuli to theuse of theKEMAR manikinwith varioustypes of complex noiseas input stimuli. Hugh Knowles served as a consultantto theVA andhis efforts, both for the VA andthe researchand developmentcommunity at large,formed the frameworkfor the evaluationprogram in place today.His contribution and leadership have resulted in proceduresthat permit the evaluation of a hearingaid in a mannervery similar to theway it will beused by the wearer. Current use of FFT analysisand strategies for evaluatingnoise reduction circuits and in-the-earhearing aids will be highlighted. 10:55 IIS. The acousticsof the external ear: Old problemsand fresh perspectives.E. A. G. Shaw (Division of Physics,National ResearchCouncil Canada, Ottawa, Ontario K IA 0R6, Canada) A coherentpicture of theexternal ear, operating both in the freefield and as an enclosedreceiver, came into focusmore than a decadeago though several troublesome problems remained, particularly at frequencies greaterthan a fewkHz. Theseincluded uncertainty about the geometry and acoustical characteristics ofthe ear canal,uncertainty about the dynamics of theeardrum and the reflection of soundfrom the eardrum, uncertain- ty aboutthe role of the externalear in soundlocalization and, aboveall, uncertaintyabout the acoustical interactionsbetween earphones and ears.Considerable progress has been made in mostof theseareas during the pastfew years but someproblems remain intractable. In the meantime,the scientificand technological advancesof the 1970sand early 1980shave providedus with a legacyof valuableinstruments: eardrum simulators,artificial heads especially the KnowlosElectronic Manikin (KEMAR), microphonesfor probing ear canalpressure, and improvedinsert earphones. Contributed Papers 11:20 and evaluated.The beamformeradapts to preservethe targetsignal (as- II6. Evaluationof a real-time adaptlve-beamforminghearing aid. Julie sumcdto bestraight ahead) and to minimizethe power of jammersignals E. Greenberg and Patrick M. Zurek (MIT Research Laboratory of arriving from all other directions.The basicGriffiths-Jim algorithm is Electronics,Room 36-761, Cambridge,MA 02139) augmentedwith a methodto inhibit adaptationin the presenceof strong target signals[Greenberg et al., J. Acoust.Soc. Am. Suppl. I 85, S26 A real-timetwo-microphone mortaural hearing aid, basedon the con- (1989) ]. The real-timesystem employs the Motorola DSP56001and was strainedadaptive beamformer proposed by Griffiths and Jim [IEEE evaluatedusing a speechtarget and a singlejammer for a varietyof free- Trans.Antennas Propag. AP-30, 27-34 (1982) ], hasbeen implemented spaceand head-mountedmicrophone configurations in anechoicand re- S87 J. Acousl.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S87 verberantenvironments. The systemdemonstrates very good perfor- A newtheoretical procedure was developed for optimizingthe acous- mance (30- to 40-dB gain from input to output in an tic characteristicsof hearingaids. The algorithm increasespredicted intelligibility-weightedspectral average of target-to-jammerratio) for speechintelligibility while keepingthe estimatedloudness of amplified severalbroadside microphone configurations in the anechoicenviron- speechat a predeterminedlevel. The intelligibilitycriterion was the mutu- ment.Adaptation inhibition provides a robustinsensitivity to targetmisa- al informationbetween spoken phonetic messages and thecorresponding lignment at high-input target-to-jammerratios. As expected,perfor- streamsof auditorypatterns in a psychoacousticmodel of the listener's mancedegrades with increasingreverberation; at the criticaldistance, the auditorysystem. The speechsignal was modeled as a sequenceof allo- systemprovides 5- to 10-dB gain. In extremereverberation, the system phonetransitions. Each transition was represented by threeauditory pat- performsat leastas well asa delay-and-sumbeamformer. [ Work support- terns,each containing ten cepstrumcoefficients, one voiceand one dura- ed by NIH. ] tion parameter.The shapeof optimizedhearing-aid frequency responses dependedstrongly on the desiredloudness of amplifiedspeech. When 11:35 normalloudness was desired for speechat 65 dB SPL, and the optimiz- ation was initializedwith the NAL prescription,hearing-aid gain was II7. Optimization of hearing-aid gain and frequency response for reduced,but predictedspeech intelligibility was nearly unchanged.Pre- eoehlcar hearing losses. Arne Leijon (Department of Information dictedintelligibility of amplifiedspeech in a noisybackground was similar Theory, Chalmers University of Technology, S-412 96 Gothenburg, for severaldifferent frequencyresponses. The mutual information was Sweden) highlycorrelated with a modifiedversion of the articulationindex. THURSDAY MORNING, 30 NOVEMBER 1989 ST. LOUIS BALLROOM H, 9:00 TO 11:45 A.M. SessionJJ. Underwater AcousticsV: The Sea Surface•Noise and Scattering Eric I. Thorsos, Chairman AppliedPhysics Laboratory, University of Washington,Seattle, Washington98195 Chairman's Introduction--9:00 ContributedPapers 9:05 Collectivemotions of bubbleclouds are experimentally investigated as an oceanambient noisesource. The experiments,performed in a water JJl. Further studiesof the underwaternoise produced by rainfall. Paul tank, show that the frequencydue to the bubblecolumn resonancede- A. Elmore, Hugh C. Pumphrey,and LawrenceA. Crum (National pendsinversely on the logarithmof the voidfraction of the water/bubble Centerfor PhysicalAcoustics, University of Mississippi,University, MS mixture.It is alsoobserved that the frequencydepends inversely on the 38677) volumeof thebubble column. These experimental results indicate that the A studyof thesound produced by waterdrops striking a watersurface volumeresonance of the bubblecolumn is one of the more likely mecha- hasconfirmed some earlier results [ Pumphreyet aL, J. Acoust.Soc. Am. nisms for an ocean ambient noise source between 100 and 1000 Hz. 85, 1518 (1989) ]. In particular,for a certainwell-defined range of drop [Work supportedby ONR and KOSEF (SWY).] a•On leavefrom the sizesand impactvelocities, drops will predictablyand repeatedlyentrain Departmentof Physics,Sung Kyun Kwan University,Republic of Korea. bubbles;this phenomenonhas been named regular entrainment. In the presentstudy, various fixed drop diametershave been used to investigate how bubblefrequency, dipole strength, and timebetween drop impact and bubbleformation vary with impactvelocity. It is found that as impact velocityis increasedthrough the point whereentrainment begins, both 9:35 frequencyand dipole strength decrease to a minimumvalue and then rise againas the highestvelocity at whichentrainment occurs is approached. JJ3. Underwatersound generated by impactsand bubbles from raindrops Both termsshow increased variability near the critical upperand lower at obliqueincidence. Armagan Kurgan, "• HermanMedwin, and Jeffrey velocities.The frequencytends to increasemonotonically as drop size is A. Nystuen (Departments of Physics and Oceanography,Naval reduced;drops that entrainbubbles at their terminalvelocities tend to PostgraduateSchool, Monterey, CA 93943) producefrequencies near 14 kHz, which is alsothe peakfrequency of the naturalrainfall spectrum.Finally, the time betweendrop impactand bub- The soundgenerated by rainfall at seais causedby raindropsof a wide ble formationwas found to increasemonotonically as drop sizeor impact size rangethat fall at their terminal velocitiesand generallystrike the velocityincreases. [Work supportedby Officeof Naval Research.] oceansurface at local, obliqueincidence. Laboratory experimentshave beenconducted to evaluatethe energyspectrum of soundscaused by both the impactand the bubbleformed by singleraindrops. The results,using terminal velocitiesand obliquetrajectories, are very differentfrom the 9:20 publishednormal incidence,nonterminal-velocity characterizations. For example,bubble frequencies other than the well-known 14-kHz peak are JJ2. An experimental investigation of bubble clouds as sources of found. Also, the energyof the impact soundincreases significantly for ambientnoise. S. W. Yoon,"} L.A. Crum (NationalCenter for Physical larger dropsand for largedeviations from normal trajectories.These ob- Acoustics,University of Mississippi,University, MS 38677), and servationsprovide specific reasons for the knownbroadening of the 14- A. Prosperetti(Department of MechanicalEngineering, The Johns kHz spectralpeak of rain noisein the presenceof windsat sea. [Work HopkinsUniversity, Baltimore, MD 21218) supportedby ONR. ] • Lt., TurkishNavy. S88 J. Acoust.$oc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S88 9:50 10:05 JJ4. Acousticalmeasurements of spectralrates of bubbleproduction JJS. ¾ertical directionality of low-frequency wind-induced noise. B. underspilling breakers. Albert (2. DanloPt and HermanMcdwin J. Softfinand W. S. Hodgkiss(P-004, Marine PhysicalLaboratory of (Physics Department, Naval Postgraduate School, Monlerey, CA the ScrippsInstitution of Oceanography,UCSD, La Jolla,CA 92093) 93943 ) The fine-scale structure of the directional ambient noise field is re- It hasbeen shown that spillingbreakers in the laboratoryproduce a vealedusing a largeaperture (900 m} verticalarray. Frequencyand spa- varietyof bubbleswhose total incoherentintensity is the Knudsensea tial spectralestimates are calculated during the passageof a localstorm, noisespectrum from I to 20 kHz [ H. Medwin and M. M. Beaky,J. providinga detailedstudy of ambientnoise levels at lowfrequencies ( 15- Acoust. Sec. Am. 86, 1124-1130 (1989)]. By usinga hydrophoneto 130 Hz) as wind speedincreases from 2 to 12 m/s over a 21-h period. simultaneouslytrigger both a "computerscope"and a spectrumanalyzer, Spectrallevels of beamsdirected horizontally show the timeand spatial it is nowpossible to positivelyidentify near-surface bubbles and to count variabilityof distantsources: Spectral levels of beamsdirected toward the their rate of productionper unit areaof the breakingwave. This passive surfacereflect local noisesources and displaya thresholdtype behavior acousticaldetermination spans a muchgreater range of bubbleradii, with with levelchanges up to 5 dB, suggestingthe abruptonset of a source a radiusresolution that is an orderof magnitudefiner than optical mea- mechanismsuch as breaking waves. Subsequent thresholds may indicate a surements[e.g., Y. Toba,Mem. Coll. Sci.,Univ. Kyoto, Ser.A, XXIX(3) changein sourcemechanism such as the conversion from spilling breakers ( 1961) ]. [Worksupported by ONR.] a•Lt., U.S.Navy. to plungingbreakers. [ Work supportedby ONT. ] 10:20--10:30 Break 10:30 11:00 JJ6. The scatteringof sound by a cylindrical bubble cloud. Murray J J8. Simulation of underwater acoustic scattering from large bubble S. Korman (PhysicsDepartment, U.S. Naval Academy, Annapolis, plumes.Stanley A. Chin-Bing(Naval Oceanographicand Atmospheric MD 21402), Ronald A. Roy, and LawrenceA. Crum (National Center ResearchLaboratory, Stennis Space Center, MS 39529) and Joseph for PhysicalAcoustics, University of Mississippi,University, MS 38677) E. Murphy(Department of Physics,University of New Orleans,New Orleans, LA 70148) Preliminaryexperiments are preformedon the scatteringof acoustic pulsesby a cylindricalbubble cloud. The goal is to investigatethe role The finite-elementocean acoustic computer models, FFRAME and playedby experimentallength scales in establishingthe angulardepend- PE-FFRAME, havebeen used to gaininsight into the dominantmecha- enceof the scatteredintensity. These length scales include bubble diame- nismsinvolved in underwateracoustic scattering by largeentrained bub- ter, interbubbledistance, acoustic wavelength, and clouddiameter. The bleplumes near the ocean surface. Various types of incidentacoustic fields bubbles(which are larger than resonancesize) aremuch smaller than the (normal mode,PE, etc.} are usedto interactwith the plumewhile using wavelengthin water,which, in turn, is smallerthan the diameterof the the finite-element models to calculate the full-wave solution for the total cloud.Experiments are carried out in a l-m3 tank using 250-kHz acoustic field.Initially, the plumeis treatedas an acousticallysoft boundary and pulsesgenerated by a 5-cm-diamcircular plane array. An identicalreceiv- variousstages in theplume's life cycle are simulated. The simulation stud- ing array detectedthe scatteredenergy as a functionof angle.The cylin- iesinclude the interaction of differentpropagating modes with thcplume, drical bubblecloud is generatedby a constantvolume velocity airflow andthe effects of frequencyand angle of incidenceon thebackscattered througha fritted ceramicdisk locatedat the bottom of the tank. Results field. arecompared with theoriesinvolving spherical clouds as well as homoge- neousfluid cylinderswith an "effective"compressibility and density. [Work supportedby ONR. ] 11:15 10.'45 &19. Acousticscattering from a "Pierson-Moskowitz"sea surface. Eric JJ7. The wedgeassemblage method applied to backscatteringfrom 3-D I. Thorsos(Applied PhysicsLaboratory, University of Washington, wind-driven sea surfaces. Richard S. Kciffer (Naval Ocean Research and Seattle, WA 98105) DevelopmentActivity Acoustic Theory and Simulation Branch, Stennis An exactintegral equation method has been used to studylow-fre- Space Center, MS 39529) and Jorge C. Novarini (SYNTEK Engineeringand Computer Systems, Inc., BaySt. Louis,MS 39520) quency(200 Hz) acousticcoherent and incoherent scattering from sur- facesthat vary in only one horizontaldirection. The surfaceroughness Numerical simulationsof acousticbackscatter from fully 3-D wind- spectrumis obtainedfrom the Pierson-Moskowitzfrequency spectrum drivensea surfaces are performedand comparedagainst experimental fora fullydeveloped sea. Comparisons are made with standard theoretical data. For this work, the directionalseas considered arc generatedfrom predictionsusing the Kirehhoff approximation, perturbation theory, and either the Pierson-Moskowitzor JONSWAP spectrumand the back- thecomposite-roughness model. The Kirehhoffapproximation is found scatteringis calculatedvia a wedgeassemblage (WA) methodthat has tobe reasonably accurate for near-specular scattering and for the coherent beengeneralized to handle3-D surfaces.The WA approachis a time- reflectionloss at grazingangles as low as 10 •. Perturbationtheory, when domainmethod that providesthe acousticimpulse response in termsof carriedto order (kh) 4 for the scatteringcross section, is in essentially diffractionsand image contributions from the scatteringsurface. The sta- perfectagreement with exact results for bistaticbackscattering. This is tisticsof the scatteredfield are calculatedby consideringan appropriate true for kh valuesas great as 1.8,where k isthe acousticwavenumber and ensembleof surfacesrealizations. Comparisons between simulations and h is the rms waveheight.The standardcomposite-roughness model is experimentaldata are examined for low frequencies( < 500 Hz) andfor foundto be accuratefor backscattering,except at the lowestgrazing an- grazingangles 10-89 deg. gles,where it over-predictsthe scattering. [Work supportedby ONR. ] S89 d.Acoust. Sec. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society ofAmerica S89 11:30 berationin thevicinity of a ship.A numericalmodel is discussed that uses asinput data bubble population densities derived from hydrodynamic and JJ10. Enbancedpropagationlnafoamymedium.A. Cowley (Admiralty bubbletransport calculations, on a three-dimensionalrectangular grid. ResearchEstablishment, Portland, Dorset DT5 2J$, United Kingdom), Acousticproperties as a functionof frequencyare computedat eachgrid C. H. Harrison (YARD, Scientific House, 40-44 Coorobe Road, New point usingclassical formulas, accounting for both resonantand non- Malden KT3 4QF, United Kingdom), A. Baird (Admiralty Research resonantcontributions. Relatively fast algorithms have been developed to Establishment,Portland, Dorset DT5 2J$, United Kindgore), T. calculatethe three-dimensionalray paths(curved or straightrays), and V. Gedrich, andC. A. Ferrari (YARD, ScientifcHouse, 40-44Coombe transmissionloss contours for arbitarilypositioned source and receiver. Road,New MaldenKT3 4QF, UnitedKingdom) Reverberationlevel as a functionof time isdetermined assuming straight Entrainedbubbles have well-known effects on the acousticproperties rays, but includingsimple multipaths and surfacebackscattering. Some of water, namelythe velocityof sound,absorption, and backscattering. interestingexamples are presentedincluding cases in whichtransmission These,in turn, havea stronginfluence on soundpropagation and fever- may be enhancedby ductingor focusingat certainfrequencies. THURSDAY MORNING, 30 NOVEMBER 1989 DIRECTORS' ROW 26, 9:30 A.M. TO 12:00 NOON Meeting of Accredited Standards Committee S12 on Noise to be heldjointly with the Technical Advisory Group for ISO/TC 43/SC1 Noise and ISO/TC 94/SC12 Hearing Protectors W. Melnick, Chairman S12 OhioState University,University Hospital Clinic, 456 Westloth Avenue,Columbus, Ohio 43210 H. E. von Gierke, Chairman, TechnicalAdvisory Group for I$O/TC 43/$C 1 1325Meadow Lane, YellowSprings, Ohio 45387 StandardCommittee S12 on Noise.Working group chairs will reporton their progressunder the plan for the productionof noisestandards. The interactionwith I$O/TC 43/SC1 andISO/TC 94/SC12 activities will also bediscussed, with referenceto theinternational standards under preparation. The Chair of the respectiveU. S. TechnicalAdvisory Groups (H. E. von Gierke) will reporton currentactivities of thesecommittees. Plans for the next meetingof ISO/TC43 and ISO/TC43/SCI, to take placein Copenhagenduring April 1990,will be discussed. S90 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S90 THURSDAY AFTERNOON, 30 NOVEMBER 1989 ST. LOUIS BALLROOM F, 1:00 TO 4:20 P.M. SessionKK. EngineeringAcoustics V: Noise Sources,Analysis, and Active Cancellation Howard H. Schloemer, Chairman Naval UnderwaterSystems Center, New London, Connecticut06320 Chairman's Introduction--l:00 ContributedPapers 1:05 1:35 KKI. Simultaneousscreech tone modesof an axially symmetricjet. KK3. The sourcesof trans-spectralcoherence in the radiation from Y. Umeda,R. Ishii (Departmentof AeronauticalEngineering, Kyoto rotatingblades. P.G. Vaidya (Departmentof Mechanicaland Materials University, Kyoto 606, Japan), and Alan Powell (Department of Engineering,Washington State University, Pullman, WA 99164),Jeffrey Mechanical Engineering,University of Houston, Houston, TX 77204- J. Girard (The Aerospace Corporation, Mail Station M4-910, El 4792) Segundo, CA 90245), and Carey L. Moulten (Wyle Research, El Segundo,CA 90245) The screechfrequency of axiallysymmetric choked jets fallssteadily with decreasingpressure ratio, with severalfrequency jumps, some up, Rotatingblades ( installedin fans,propellers, and helicopters) radiate somedown [A. Powell,Proc. Phys.Soc. London Ser. B 66, 1039-1056 both discrete and broadband sound. This radiation can be examined for (1953)]. Daviesand Oldfield's Lissajous figures [Acustica 12, 267-277 thecoherence of thetrans-phase. Trans-phase isthe phasedifference be- (1962) ] showedthat the correspondingjet oscillationmodes are of var- tweenthe componentsat two differentfrequencies when one of themhas ioustypes. Schlieren photography and digitally processed acoustic pres- beensent through a nonlinearfilter. If thetrans-phase shows coherence, suresignals are now used. Unambiguously, the first two modes, a• anda 2, the signalsare saidto betrans-spectrally coherent. Various sources of the are axiallysymmetric (varicose, like ring vortices);the next,b, is asym- rotatingmachinery noises contribute to the coherencein differentways. metricin anarbitrary plane (sinuous); c ishelical, rotating in eithersense, Existenceof suchcoherence can help in thediagnostics and identification stableonce started, while d is sinuousagain, connected to c by hysterectic of thesources and in thedetection and classification of the machinery in jumps. However,tone d is weakly but smoothlyconnected to b, so d presenceof extraneousnoise. appearsto be an extensionof b. Moreover, the dominantc tone extends weaklyto overlapwhere b and d dominate,and a suspectedweak a2(?) overlapswhere b, c, and d dominate.Further, both a• and b existweakly 1:50 wherea 2 dominates.Evidently jumps occur not necessarilyat changesin themode as previously thought, but whenthere is a changein whichone KK4. On the feasibilityof the evaluationof the performanceof a plnger dominates. transducerin an opentank usinga pressurizedvessel. V. N. Bindal, T. K. Saksena,S.C. Gupta, and D. R. Cbaubey (National Physical Laboratory, New Delhi-12, India) Investigationon the calibrationof pingertransducers to a hydrostatic pressureof 200 psiis described.The transduceris housedin a specially constructedvessel fabricated from fiberglassin polyesterand epoxy ma- 1:20 trix of 10-ram wall thickness. The insertion loss of the walls of the vessels hasbeen estimated in thefrequency range 10-20 kHz andis observed to be KK2. Recentexperiments on impingingchoked jet tones.Alan Powell small.Source level measurements have been made by puttingthe pressur- and BrendaHenderson (Department of Mechanical Engineering, ized vesselin a tank of dimension6.35 mX5.85 mX2.2 m3. The merits Universityof Houston,Houston, TX 77204-4792) anddemerits of the vesselsusing an epoxyand polyester base have been Two modesof oscillation,L andS, associatedwith largeand small discussed.Measurements show that thereis no significantvariation in plateswere described, with frequencyanalysis by counter[A. Powell,J. acousticoutput of thepinger transducer up to 200 psi,although a vari- Acoust.Soc. Am. 83, 515-533(1986) ]. With differentslopes on log-log ationis observedin the frequencyresponse characteristic. The methodis plotsof wavelengthversus nozzle-to-plate spacing h, feedbackto nozzle suitablefor characterfzation ofa transducerat high pressures that cannot andto stand-offshockwave,respectively, was hypothesized. Digital signal be characterizedby usingconventional techniques. processingis morerevealing! The samedominant and S toneare found, exactly.It nowappears these L and$ arethe dominant members of classes of tones.At smallh, for largeplates (all pressures)and small plates (low 2:05 pressures),L tonesare observed in multiplemodes, usually simultaneous- KKS. Accurate spectral estimationof multiple sinusoidsusing an FFF. ly. For intermediateh, frequencyjumps occur;schlieren photography JohnC. Burgess(Department of MechanicalEngineering, University of revealssymmetric and asymmetricjet flow oscillations.Nozzle reflector Hawaii, Honolulu, HI 98622) influencesupports the feedback to the nozzlehypothesis for L. For largeh at low pressures,three strong constant tones exist--screech a• and a2- Accurateestimates of amplitude,frequency, and phaseof multiple and oneother(?). At higherpressures (small plates), S tonesappear, sinusoidsin a signalcan be obtainedusing a methoddescribed earlier [J. influencedby a nozzlereflector (!); schlierenphotographs exhibit sym- C. Burgess,J. Acoust. Soc. Am. Suppl. 1 83, S92 (1988)]. FFT size metrical oscillationsof jet and stand-offshockwave. For intermediate requiredfor specifiedspectral resolution, as well as somecomputational plates,both L and S classesoccur. aspects,is discussed. S91 J. Acoust.Soc. Am. Suppl.1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S91 2:20 3:05 KK6. Low peak factor waveformsfor efficient systemanalysis using a KK9. Sound-pressuredistribution on a crosssection of a duct for active fast Fourier transform.John C. Burgess(Department of Mechanical noisecancellation. Kang Yen, OsamaA. Mohammed,Vijay Raman, Engineering,University of Hawaii, 2540 Dole Street, Honolulu, HI and Kurt S. Ramdin (Department of Electrical Engineering,Florida 96822) InternationalUniversity, Miami, FL 33199) Low peakfactor waveforms [M. R. Schroeder,Number Theoryin Activecancellation techniques for low-frequencynoise have been ap- Scienceand Communication(Springer, New York, 1984)] can be de- pliedto someindustrial plants. Field testsshow very useful results. In the signedso that multipleharmonics in the waveformappear to havethe past,research work has concentrated on the theoryof minimizationof the idealspectrum of a trueFourier transform when the waveform is analyzed enclosedsound field along one axis. However, in practicethe directionof usinga fastFourier transform (Fl:rr). If the harmonicsin the waveform the soundsource is uncertain.Especially when the noisesource is very arerestricted to thosehaving an integralnumber of cyclesin theFFT data closeto theduct, the direction of thesound wave will bea dominantdesign recordlength, there is no leakage,and the magnitudesand phasesof the factor in order to achieve the cancellation at the other end of the duct. In Fourier coefficients are identical to those of the harmonics in the wave- this paper,a setof directionaland finite-sizecancellation sources inside form. Low peakfactor waveforms drastically reduce peak power require- the duct is considered.The sound-pressuredistribution across any cross ments.For example,the peakpower requirement for a waveformhaving sectionin the duct is the primary researchinterest. The resultsprovide 63 harmonicscan be reducedby a factor of about 30. Waveformscan be informationon the designof noisecancellation devices such as noise can- designedhaving harmonics with constant,octave, and closeapproxima- cellationearphones. [Work supportedby the Florida High Technology tionsto one-thirdoctave frequency differences. This approachcan pro- and Industry Council.] videan efficientalternative to othermethods in currentuse for measuring systemtransfer functions. [Work supportedin part by the NationalRe- searchCouncil of Canadaand the Lanikai Foundation.] 3:20 KKI0. Active attenuationof pure tonesin a lined duct usingadaptive 2:38 filtering. Dominique Laget and Elmer L. Hixson (Department of Electn'baland ComputerEngineering, University of Texas,Austin, TX KK7. Application of the fast Hartley transform to industrial noise 78712) analysis.Woon S. Gan (AcousticalServices Pte Ltd., 29 Telok Ayer Street,Singapore 0104, Republic of Singapore) Fan noiseis characterizedby strongdiscrete blade passing tones add- ed to broadbandaerodynamic noise. In air conditionerducts, absorptive The applicationof thefast Hartley transform (FHT) to industrial liners provide reductionof the noiseat the higher frequenciesbut are noiseanalysis to replacethe fastFourier transform (FIT) is proposed. ineffectiveat lower frequencies.An activeadaptive controller is mostly The Hartley transformof a real-valuedfunction is alsoreal-valued. Its effectivefor low-frequencynarrow-band noises. In this study,noise con- evaluationdoes not involvecomplex functions and it has the samefor- trol combininga duct liner and an adaptiveattenuator has been imple- mula for forwardand inversetransform. Thus it hasa speedadvantage mented.The stabilityand convergence properties of theadaptive attenua- overFFT. The typeof FHT algorithmchosen is dueto H. V. Sorensenet tor dependon the compensationof the transducersand acousticdelays. al. [IEEE Trans.Acoust. Speech Signal Process. ASSP-33, 1231-1238 Compensationsusing an IIR filter, a delay,and an FIR filter designed ( 1985) ]. As an illustration,the two-channel FIT signalprocessor ischo- with the LMS algorithmhave been studied. After simulation,the control- sen.The computationof the followingfunctions is repeated:autocorrela- lersbased on a filtered-xand a delayed-xLMS algorithmhave been imple- tionfunction, cross-correlation function, power spectrum, and the acous- mentedon a digital signalprocessing board. Measurementsinside and tic intensityof real-numbereddata by replacingthe FIT with FHT. A outsidea lined air conditionerduct showedsignificant pure-tone attenu- detailedanalysis of the time that canbe savedin computationby using ation, especiallywhen a novelfiltering-x method is used. FHT is made.A key advantageof the Hartley transformtechnique for largeautocorrelation is the useof the sameformula for transformingand inversetransforming. 2:50 3:35 KKS. Finite element simulationof an active noise-cancelingscheme. KKll. Numerical simulationof a soundfield generatedby flow arounda OsamaA. Mohammed,Kurt S. Ramdin,Vijay Raman,and Kang Yen circularcylinder. Z. C. Zhu andL. Z. Zeng(Department of Engineering (Departmentof ElectricalEngineering, Florida International University, Mechanics,Tsinghua University, Beijing, People's Republic of China) Miami, FL 33199) Flowaround a circularcylinder is a classicalproblem, but it isdifficult This paperdetails a finite-elementcomputer simplation scheme of an to simulatea Karmanvortex street in the wakenumerically. In previous axis-symmetricphysical setup of an active noisecanceling headset. The numerical works artificial rotation had to be added to the cylinder to resultsof this simulationshow the distributionof sound-pressurelevels separatethe vortex from it. Obviouslyit would influencethe original insidethe car canal.Two-dimensional simulations of an axis-symmetric solution.In addition,body conditions were usually not satisfiedcorrectly configurationof the ear cupwere carried out. A pointsource was used to in many numericalworks. There are two major improvementsin this cancel out the noise. From these simulations it can be concluded that the paper. ( 1) In order to separatethe vortex,it is proposedthat only one noise inside the ear canal can bc reduced. The amount of reduction de- asymmetricalgrid-point condition is givenat thefarfield of the wake. (2) pendson the strengthand phaseof the cancelingsource. On the other A recenttreatment is usedto guaranteethat bodyconditions are satisfied hand,a pointsource is not a practicaldevice. To overcomethis problem, precisely.Then numericalsimulation of flow arounda circularcylinder new structuresfor speakershave to be createdand simulatedto checkif canbe completed correctly and successfully, with theapplication of ADI- their radiationis in accordancewith that of practicaldevices. An algo- BGE schemesto solveincompressible N-S equationsin the form of rithm proposedearlier by theseauthors for the estimationof the SPL noise streamfunction and vorticity at Reynoldsnumbers from 1 to 1000.Sound reductionthat canbe accomplishedinside the ear canalhas been added. radiationis calculated by numericalintegrating FW-H equationsat anRe [Work supportedby the Florida High Technologyand IndustryCoun- of 100. Then carefulanalysis and comparisonare madefor all partsof cil. ] soundsources. Finally severaluseful conclusions are given. S92 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S92 3:50 stable.It isalso time-saving and storage-saving. Maybe it isa goodmethod for calculatingshocked solutions of soundpropagation in ducts. KK12. Numericalanalysis of soundpropagation with acousticshock wavesin varyingarea duets. Z. C. Zhu andY. Zhang(Department of EngineeringMechanics, Tsinghua University, Beijing, People's Republic 4:05 of China) KK13. The physicsmodel of thyristor interferencein radios and other In recent results,several second-orderfinite differenceschemes were communicationequipment. Ming-Fu Yang, Fci Xue, and Yi-Pei Lu adaptedto obtainshocked solutions for soundpropagation in varyingarea (No. 357 Xun Yang New Village, Shi Quan Road, Shanghai200061, hard-wallducts with transonicflow [ N.J. Walkingtonand W. Eversman, People'sRepublic of China) J. Sound Vib. 104, 81-107 (1986)]. The second-orderMacCormack schemewith a fourthviscous term was then recommended by theauthors. With the increasinguse of thyristorsin electricinstallations such as But its numericalerror still seemedto be a bit large.In thispaper, the lamp controllers,electric fan controllers,and electricheat controllers, fourth-order MacCormack scheme with a fourth viscous term is used to thyristorinterference in the middlefrequency (.f< 20 MHz) rangeis re- improvethe solution.It is shownfrom threesample calculations with pugnantand seriously hampers the useof thyristors.A physicalmodel of acoustic shock waves that the new method is much better than the second- thyristorsindicates that interferenceresults from the instantaneouspulse orderMacCormack method. Its numericalerror, on an average,is only that is producedwhen the thyristoris turnedon. Suchinterference will onefifth of thatin thelatter method. Here the new method is depicted in producean interferencepulse on a nearbyantenna that is receivedand detail,considering especially the treatmentof boundaryconditions, the amplifiedto becomeannoying noise. This conclusionis usefulin the de- selectionof viscouscoefficient, the stability condition, and the computa- signof electricinstallations using thyristors so that suchinstallations have tioncorrectness criterion. This method needs less grid points and is more greaterreliability and better interferenceprevention. THURSDAY AFTERNOON, 30 NOVEMBER 1989 DIRECTORS' ROW 43, 1:00 TO 3:15 P.M. SessionLL. Physical AcousticsVI: Ultrasound and Nondestructive Evaluation Allan D. Pierce, Chairman GraduateProgram in Acoustics,Pennsylvania State University,State College, Pennsylvania 16804 ContributedPapers 1:00 cousliquid to a solidlayer is observed.When the interfacelayer is in the liquid statewith very smallshear nodulus the shearstress components on LL1. Elastic constantreconstruction using resonances of coupledmodes the boundaryvanish, resulting in a slipboundary condition. On theother in thin anisotropicplates. S. I. Rokhlin, C. Y. Wu, and L. Wang hand,after thelayer is solidified,the two semispacesbecome welded with (Departmentof WeldingEngineering, The Ohio State University,190 thetransition of shearand longitudinal stresses to oneanother. It isshown West 19th Avenue, Columbus, OH 43210) that the ratio of a hydrodynamicboundary layer thickness to the interface At obliqueincidence of ultrasonicwaves on a thin anisotropiccom- layerthickness plays a majorrole in thistransition. A simpleapproximate positeplate immersedin fluid, the quasilongitudinal,flexural, and quasi- equationfor the reflectioncoefficient from the interfaceis derivedand SH vibrationsare excited.At someangles of incidence,coherent interac- comparedto the exactsolutions for differenttransitional stages of the tions of thesewaves lead to minima of the normal displacementon the boundarycondition. Theoretical predictions are comparedto experimen- back surfaceof the plate and thereforeto minima of the transmission tal results. coefficient.At off-axisorientations of the thin compositeplate, two such minima may be observed.Their positionsare only slightlydependent on frequencyand are determinedby plateproperties and orientation.This effectwas studied both theoretically and experimentally.The elasticcon- 1:30 stantshave been reconstructed for a unidirectionalgraphite-epoxy com- positeplate from experimental data that havebeen taken at differentplate LL3. Method for measurementof elastic constantsof thin anisotropic orientationsand frequencies. membranes.W. Wang and S. I. Rokhlin (Department of Welding Engineering, The Ohio State University, 190 West 19th Avenue, Columbus, OH 43210) In thiswork theelastic properties of a thin layerof porousaluminum 1:15 oxide obtainedby phosphorusanodization have beenstudied by a novel LL2. On the transition betweenslip and rigid boundaryconditions ultrasonictechnique. The thicknessof the porousoxide samples is about betweentwo solid media. Y. J. Wangand S. I. Rokhlin(Department of 50,urnwith through vertical pores of about 2000 • diameterand porosity Welding Engineering,The Ohio StateUniversity, 190 West 19th Avenue, of about70%. The ultrasonictechnique is basedon measurementsof the Columbus, OH 43210) angleof the minimum of the transmittedenergy at obliqueincidence on the immersedsample. The effectiveelastic constants that havebeen theo- In order to describethe transitionbetween slip and rigid boundary reticallycalculated from the experimentaldata are in goodagreement conditions,a modelwith a thin interphasialviscoelastic layer is intro- with thatfound from estimations of bulkproperties of anisotropicporous duced.By changingthe relaxationtime (to•'), the transitionfrom a vis- media.The knowledgeof the propertiesof porousanisotropic anodized S93 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S93 Al:03 layershas major importancefor adhesivejoint characterization known.Convergence and stabilityof the methodwill be discussed,and sincesuch layers play an importantpart in interphasialstrength of the examplesof field calculationsfor a few simpleprofiles will be given. joint. [Work supportedby Centerfor NDE at Iowa StateUniversity.] 1:45 2:30 LIA. Focusingof ultrasoundin solidswith applicationsto QNDE. John LL7. Surfacemotion reconstruction based on half-space field data. David G. Harris (Center for Quality Engineeringand Failure Prevention, D. Benninkand Anna L. Pate(Department of EngineeringScience and CatalysisBuilding, 2137 North SheridanRoad, Evanston,IL 60208- Mechanics,Iowa StateUniversity, Ames, IA 50011) 3020) A finitebut otherwisearbitrary normal velocity profile at the bound- Ultrasoundcan be focusedby the curvedfluid-solid interfaceof a ary of a half-spaceis expandedin a setof orthonormalfunctions. The solidpart being interrogated for cracksor by usinga lensas part of an radiationproblem is then formulatedin terms of the momentsof this imagingsystem such as the acousticmicroscope. In the lattercase it is profile.(Details of thisformulation are given in thepreceding abstract.) usuallydifficult to forma diffraction-limitedspot when the focuslies in The radiationproblem is discretizcd to obtaina matrixformulation and a the solid becausethe focusedbeam field must crossa fluid-solid interface. singularvalue decomposition method is used to estimate the velocity pro- This papersummarizes recent investigations of thesetwo situations.A file from measurementsof the radiated field. Additional constraints are calculation of the wave fields near the eaustiesformed when a two-dimen- imposedupon the derivativesof the profilein orderto overcomethe ill- sional (unfocused) beam scattersfrom a concavefluid-solid interfaceis posednature of the inversesource problem. The accuracyof the method discussed. Also discussedis a calculation of the wave fields within the will be discussedin termsof boththe positioningof the fieldlocations focalregion of a three-dimensional,wide aperture source when the fo- (nearfieldversus farfield) and the weighting of constraints.Examples of cusedbeam is directedacross a planefluid-solid interface and the focal velocityprofile reconstruction will be givenfor a few simpleradiators. regionlies in the solid.In both casesattention is paidto the interplay [ Worksupported by Center for NDE at IowaState University. ] betweenthe diffraction effects of the sourceand of thefocal region, and alsoto the excitationof leakysurface waves. The implicationsof these considerationsfor the quantitativeinterpretation of ultrasonicmeasure- 2:45 mentsare explored. LLS. A prior knowledgebased optimal Wiener filtering approachto ultrasonicscattering amplitude estimation. Steven P. Ne• (Mechanical and Aerospace Engineering, University of Missouri---Columbia, 2:00 Columbia,MO 65211) and Donald O. Thompson(Center for NDE, Iowa StateUniversity, Ames, IA 50011) LLS. Guidedwave generation by directexcitation. Peter B. Nagyand LaszloAdler (Departmentof WeldingEngineering, The Ohio State In orderto interpretmeasurements of the scatteredwave field result- University,Columbus, OH 43210) ingfrom sonification of anobject, the effects of themeasurement system must be removed from the measured data. In ultrasonic nondestructive Guidedacoustic waves along interfaces and more complex layered evaluation,estimation of the farfieldscattering amplitude for a flawin a structuresare often used in ultrasonicNDE. Theiroutstanding sensitivity material,e.g., a crack,void, or inclusion,involves removing the measure- to specificproperties associated with interface conditions and bond quali- mentsystem effects in thepresence of electronicnoise and noise due to the ty isdue to thefact that their energy is effectively confined to theregion of scatteringof soundfrom inherent material features such as grainboun- interest.On theother hand, this advantage turns out to bea greatdraw- daries.The purpose of thework reported here was to evaluatean optimal backfor generationand detectionof suchguided waves. Certain modes Wienerfiltering approach to scatteringamplitude estimation. The filter with higherphase velocity than at leastone of the bulk velocitiesof the wasshown to determinean optimalestimate as the weighted average of surroundingmedia are called leaky modes that can be coupled to viaphase the information derived from measurement of the scattered wave field and matchingbulk waves. True guided modes of lower phase velocity than any priorinformation about the flaw distribution. The optimal Wiener filter of thebulk velocities of thesurrounding media are much more difficult to wasev•.luated and comparedto a nonoptimalform of the Wienerfilter. generateand detect. A simpleexperimental technique based on longitudi- Theoptimal Wiener filter was shown to provideimproved scattering am- nal or shearwave excitation along the interfaceis shownto be veryeffec- plitudeestimates by filteringout the noiseand by utilizingprior flaw tive in generatingboth symmetricand antisymmetric guided modes of information.[ Work supportedby the Directorof EnergyResearch, Of- Rayleigh(SV) or Love(SH) typeof polarization.Experimental results riceof BasicEnergy Sciences. ] from differentplates and adhesiveand solid-stateinterfaces are shownto be in good quantitativeagreement with theoreticalcalculations. This techniquemight find numerous applications in basicacoustic studies as 3:00 well as ultrasonicnondestructive evaluation. [This work wassupported by the Officeof Naval Researchunder Contract N0014-C-2129. ] LL9. The measurementand analysis of acoustic noise as a random variable in ultrasonic nondestructive evaluation. Steven P. Neal (Departmentof Mechanicaland AerospaceEngineering, University of Missouri---Columbia,Columbia, MO65211) andDonaldO. Thompson 2:15 (Centerfor NDE, Iowa StateUniversity, Ames, IA 50011) LL6.Radiated fields determined from moments of surface velocity. Soundscattered from inherentmaterial features such as grain boun- David D. Bennink and Anna L. Pate (Department of Engineering dariesbecomes a sourceof acousticnoise when attemptingto detector Scienceand Mechanics,Iowa State University, Ames, IA 50011) characterizea flaw in a material,e.g., a crack,void, or inclusion.In formu- latinga stochasticmodel to describea flawdetection or flawcharacteriza- A quickand efficientmethod is derivedfor calculatingthe fieldradiat- tion experiment,acoustic noise is often assumedto be an uncorrelated, ed by a finitebut otherwisearbitrary velocity profile. The methodcalcu- Gaussianrandom variable with zero mean and known averagepower latesthe pressurefield radiatedinto the half-space(z> 0) from the nor- spectrum.The work reportedhere focuses on the measurementof acous- mal componentof velocityon the boundaryplane (z = 0). The derivation tic noiseand on the evaluationof theseassumptions for the casesof scat- startsfrom the Rayleighintegral and utilizesa powerseries expansion of tering from grainsin stainlesssteel and scatteringfrom porosityin alumi- the kernel in the radial coordinateof the sourcepoint. As a result, a num, respectively.An estimate of the average power spectrum is formulationis obtained that containsseparate terms for the sourceand for determined for each case. It is shown that in both the time domain and in propagation.The sourceterms represent the moments of thevelocity pro- the frequencydomain the acousticnoise considered has zero mean and is file. The propagationterms can be obtainedfrom a recurrencerelation reasonablyuncorrelated and Gaussian. [ Work supportedby the Director startingfrom certainspecial cases for which analyticalexpressions are of EnergyResearch, Offee of BasicEnergy Sciences. ] $94 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S94 THURSDAY AFTERNOON, 30 NOVEMBER 1989 ROSE GARDEN ROOM, 1:00TO 3:50 P.M. SessionMM. PsychologicalAcoustics VIII andPhysiological Acoustics IV: GeneralPhysiological and Psycho- logicalAcoustics (Poster Session) Robert H. Gilkey, Chairman CentralInstitute for theDeaf, 818 SouthEuclid Avenue, St. Louis,Missouri 63110 All posterswill bedisplayed from 1:00 to 3:50p.m. To allowall contributorsan opportunityto seeother posters,contributors of odd-numberedpapers will beat theirposters from 1:00 to 2:25 p.m. and contributors of even-numberedpapers will beat their postersfrom 2:25 to 3:50p.m. ContributedPapers MMI. Behavioralthresholds for tone pipsin 3- and 6-month-oldinfants To determinewhether the form of psychometricfunctions for 2I, and adults. Lynne Werner Olsho, M. Patrick Feehey, and Richard 2AFC leveldiscrimination changes with hearingimpairment, measure- C. Folsom (Departmentsof Otolaryngologyand Speechand Hearing mentswere obtained in listenerswith cochlearimpairments and normal Sciences,University of Washington,W J-10, Seattle,WA 98195) listenerswith impairmentssimulated by masking.Ten incrementlevels were presentedin randomorder within blocksof 100 trials. Stimuli were Infantbehavioral thresholds for tone-pip stimuli at 1,4, and8 kHz ( 3- tonesat 0.25, 1, 8, or 14kHz or whitenoises chosen to encompassa wide cyclerise-fall and 1-cycleplateau) were estimatedusing the observer- rangeof conditionsand difference limens. Durations were either 10or 500 basedpsychoacoustic procedure (OPP). Behavioralthresholds for these msand levels ranged from 50 to 90 dB SPL. For eachcondition, at least20 stimuliwere also obtained on a groupof normal-hearingadults. Across blockswere presented in mixedorder. Results show that the sensitivityd' frequencycomparisons revealed that 6-month-oldinfant threshold curves isnearly proportional to &L( = 20 log [ (p + Ap)/p], wherepispressure} paralleledadult curves,but infant thresholdswere elevated 20 to 30 dB. over the entire rangeof differencelimens for both simulatedand real For 3-month-oldinfants, OPP thresholdcurves were not parallelto adult impairmentsas well as for normallisteners tested in the quiet.These curvesdue to relativelypoorer infant thresholdsat 8 kHz: Furthermore, resultsindicate that if the transformationfrom stimulus intensity to deci- in comparisonwith waveV auditorybrain-stem response (ABR) thresh- sionvariable is affectedby heatingimpairment or masking,the change olds for the same stimuli, 3-month-old infant behavioral thresholdsare affects both the mean and the standard deviation in the same manner. the same as ABR thresholdsat 1 and 4 kHz, whereasbehavioral thresh- They alsolend support to the notionthat A/_,•plotted on a logarithmic oldsare poorerat 8 kHz. A similarelevation in thresholdat 8 kHz for 3- scale--is an appropriaterepresentation of level-discriminationperfor- montholds has been reported for longerduration stimuli [ L. W. Olshoet mance.[Work supportedby NIH.] al., J. Acoust.Soc. Am. 84, 1316-1324(1988)]. The resultssuggest a frequency-dependentmaturation of behavioralthresholds between 3 and 6 monthsfor thesebrief stimuli. [Work supportedby NIH.] MM4. Electrotactile encodingand recognitionof a 16-obstruentsystem. H.G. Piroth, S. Gfroerer, and H. G. Tillmann (Institut ffir Phonetik MM2. Centralprocessinginpatientswithcochlearimplants.D M. Daly und SprachlicheKommunikation, Schellinstrasse3, 8000 Munich 40, (Box 210855, Dallas, TX 75211}, F. D. Owens, and P. Stephenson Federal Republicof Germany) (3600 Gaston Avenue, Suite 1103, Dallas, TX 75246) In continuationof a long-termtraining experiment [H. G. Tillmann Becauseof locationsand orientationsof auditory cortex, studiesof and H. G. Piroth, J. Acoust.Soc. Am. Suppl. 1 84, S45 (1988)] the centralprocessing in mantypically exploit psychophysical phenomena or recognitionof 16 obstruentsin Co syllableswas testedusing a modified diseasealtered structure/functioning. Cochlear implants afford a viewof quasiarticulatorycoding method. Tactile syllable patterns were delivered centralprocessing through sparse, specifiable peripheral transfer func- by the 16-channelelectrocutaneous system "SEHR-2" to 16 electrode tions.Six patients were examined who have unilateral implants (cochlear pairsthat werearranged to build four equidistantrings around the left 22 channel;14-21 active electrodes in BP + 1;4 AS, 2 AD implants;2- forearm.Consonants were assigned to ringsor pairsof ringsin order to 100 weekspost-op) usingprerecorded sets of BW, GY, BDG, and yieldmore than four placesof articulation.Fricatives were coded as sub- delivereddirectly through the processor[Daly et al., J Neurophysiol. sequentstimulations of four or eightelectrodes (forming one or two tings, 140 (2), 141-62 (1980) ]. The representationof stimuliwas sufficient for respectively)while plosiveswere representedby oscillatingmovements three patientsto classifystop/glide sets consistently (p<0.001); al- betweenthe neighboringelectrodes of a ring.The fortis/lenisdistinction thoughin a setwhich included both BW andGY nonedistinguished [be] wascoded by differentpattern velocities. Since recognition with theinitial from [gel or [we] from [ye]. Performancereverted to chancewhen the codingof the two-tingfricative patterns as paralleland synchronous processorwas configured as a singlechannel device and all stimulihad the movementswas poor, an alternativecoding method was applied with sametotal duration.None distinguished among BDG. Onepatient classi- contrarotatingmovements in order to avoid local maskingor inhibition fiedvowels ile consistently;boundaries changed significantly when a cen- resultingfrom the stimulationof adjacentelectrodes. Using this alterna- tral electrodemalfunctioned and wasinactivated in configurationmap. tive coding method, one highly trained subjectreached 85.6% correct Resultssuggest, with thesestimuli, patientsresolved brief temporal identifications of the 16 obstruents. changesmore accurately than brief frequencychanges alone. MMS. Consonanttesting of a tactile hearingaid basedon amplitude-and MM3. Psychometric functions for level discrimination in simulated and time-quantized speech. Edward M. O'Brien (Bioengineering Program, cochlear impairment. Sq•renBuus (Communication and Digital Signal Texas A&M University, College Station, TX 77843-3120) ProcessingCenter, 409 DA, NortheasternUniversity, Boston, MA 02115), Mary Florentine, and Tilmann Zwicker (Communication Speechwas first prefilteredwith a second-orderbandpass filter cen- ResearchLaboratory, 133 FR, NortheasternUniversity, Boston, MA teredat 1.5kHz. Speechwas then amplitude quantized to twolevels based 02115) on the zero crossingsof the waveform.This signalwas then time quart- S95 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S95 tized such that it could only changestates in associationwith a clock MMS. Evidence for a tight blood-labyrinth barrier. Naoki Inamura, signal.The intelligibilityof the processedspeech remains high for clock AlecN. Salt,and Ruediger Thalmann (Department of Otolaryngology, frequenciesabove 10 kHz (which was the quantizingrate usedfor this WashingtonUniversity, St. Louis,MO 63110) study).The discretedigital frequencies in the resultingwaveform were easilydetected with a digitallogic circuit. These were then displayed onto Theentry of a smalltracer ion (FW 135.6)from blood into perilymph the skinof the abdomenwith a lineararray of concentricring electrodes. and cerebrospinalfluid (CSF) has beenmonitored using ion-selective Thirty-one electrodes(representing frequencies between 151 and 2500 electrodes.The tracer,trimethylphenylammonium (TMPA) can be de- Hz) wereused for thearray. Biphasic constant current pulses were used as tectedin extremelylow concentrations ( < I pro) by electrodes,allowing the stimuli.Five normalhearing college age students served as the sub- entry to be accuratelyquantified. Electrodes were sealed into the cochlea jects.Discrimination of consonantpairs was used to testthe ability of the to ensurethat artifactualfluid leakages did notdistort the results. A con- tactileaid to transmitspeech information to the subjects.The following stantlevel of TMPA in plasma(approximately 0.5 mM) wasmaintained rangesof accuracy(corrected for guessing)were observed:stops and by continuousvenous infusion of tracer.TMPA entry into CSF wasex- constituents,18% to 87%; voicedand unvoiced consonants, 3% to 27%; tremely slow, reachingonly 13% of the plasmalevel within 90 rain consonantswith differentplaces of articulation,27% to 90%; nasal, (n = 3). Tracer enteredperilymph even more slowly,reaching 6.5% glides,and semivowels,17% to 49%. (n = 5) and 3.7% (n = 5) of the plasmalevel in 90 rain for ST and SV, respectively.The slowentry of sucha smallion into perilymphdemon- stratesthe existenceof a tight barrierbetween blood and perilymph, and eliminatesthe possibility that perilymph is generated by mechanisms such as ultrafiltration.In two animalswhere bleeding was noted at the elec- trodeinsertion site, a muchmore rapid rate of tracerentry was recorded, probablyas a resultof plasmaleakage into perilymph.Treatment with epinephrine(giving a briefblood pressure increase) also opens the bar- MM6. Noise reduction hearing aids: A tutorial. Ruth A. Bentler and rier, allowingrapid tracerentry. BrendaB. Bergman(Department of SpeechPathology and Audiology, Universityof Iowa, Iowa City, IA 52242) In thelast few years, there has been a proliferationof (so-called)noise MM9. Action potentialtuning curvesrecorded with a derivedresponse reductionhearing aids onto the market.While someof thesedesigns are technique. Philip Garcia and Alec N. Salt (Department of little morethan multiband compresson circuits, others employ adaptive Otolaryngology,Washington University, St. Louis,MO 63110) filtering(dependent upon overall input level ), adaptivecompression (de- A numberof laboratoriesuse tonal maskingmethods in conjunction pendent upon duration of high-levelinput signals), and variations with the cochlearaction potential (AP) to generateAP tuningcurves. In thereof.The purposeof thisposter session istwofold: ( 1) usinggraphics thesestudies, the amplitudeof AP elicitedby a probestimulus is moni- and schematicsan explanationof six available"noise reduction" circuits toredas the levelof maskeris varied.An amplitudereduction (AR) by a will be provided.This will serveas a tutorialfor thoseaudiologists at- specifiedamount (25%-100%) is usedto indicatemasking of the re- temptingto keepup with designand marketing strategies for noisereduc- sponse.An alternativemethod was investigated in whichinteraction of tion; (2) preliminaryresults of a largeclinical project begun late in 1988 theprobe and masker is monitored by digital subtraction of the unmasked aimedat determininghow effective different circuits are for differentcon- andmasked responses, generating a tone-derived(TD) response.An am- figurationsand/or degreesof hearingloss in new and experiencedusers plitudecriterion of 10-pVTD responseis usedto indicatemasker/probe will be presented.Success of effectivenesswill eventuallybe measured interaction.Tuning curves were recorded from the round window of anes- overa 12-monthperiod using (a) speechrecognition (in quietand noise), thetizedguinea pigs using AR andTD methods.With nearthreshold (b) portionsof the Hearing PerformanceInventory (Giolas and Owens, probestimuli, similar tuning curves were obtained. With higherIcyels 1979), (c) expectationschecklist (Seyfried and Anderson,1989), (d) however,markedly different results were observed.AR tuningcurves qualitativemeasure task (modified from Gabrielsonand Lindstrom, becameelevated and sharper while TD curvesbecame lower and broader. 1985), as well as (c} ratingsof"satisfaction" taken at 6 monthsand 12 The TD methodappears to betterrepresent the frequencyspread of re- monthspost fitting. sponseswith increasing stimulus level as demonstrated at the single unit level.In addition,TD tuningcurves were able to detectsecond harmonic distortionsof theprobe stimulus, while AR tuningcurves did not. [ Work supportedby Programproject grant P01 NS24372and Training Grant NS 07278. ] MMI0. Codingof phasestructure in complexsounds by bullfrog MM7. Regressioneffects in magnitudejudgments of line length, and auditory-nerve fibers. AndreaMegela Simmons and Michael vibrotactilesine and pulse loudness.Amy A. Collins and Roger Ferragamo(Department of Psychology,Brown University, Providence, W. Cholewiak (Department of Psychology,Green Hall, Princeton RI 02912) University, Princeton,NJ 08544-1010) Thisexperiment explored how the peripheral coding of thespectral Fiveobservers performed absolute magnitude estimation (AME) and andtemporal structure of a complexcommunication signal is affected by absolutemagnitude production (AMP) of line lengthsand vibrotactile the phaserelations between individual frequency components in the pulseand sine loudnesses as well asmatching of eachpossible pair. Line sound.Complex sounds with 21components, all of whichwere harmonics lengths,presented visually, could extend over a 7- to 950-ramrange. Vi- of a common,low-frequency fundamental ( 100 Hz), weredigitally syn- brotactilestimuli were 32-ppshayersine pulses and 230-Hz sinusoidspre- thesized.The frequencies and amplitudes of theindividual components sentedto the palm with 7-mm diametercontactors and staticsurrounds. matchedthose of the bullfrog'sspecies-specific advertisement call. For Whenthe AME andAMP datawere plotted together, perfect correlation onestimulus, the 21 components were generated in cosinephase, and for was not seenindicating the influenceof regressioneffects on the data. theother stimulus, the 21 componentswere in randomphase. Most am- Regressioneffects occur when observers making judgments on two per- phibianpapilla (AP) andbasilar papilla (BP) fiberssynchronized to the ceptualcontinua (i.e., line magnitudeand numbermagnitude) restrict envelopeonly of thephase-coherent stimulus, and the complex spectral therange of theirresponses gravitating toward the mean response. Magni- structureof the stimulus was not accurately encoded by population firing tude balance functions (MB) were calculated from AME and AMP data patterns.Envelope synchronization to the random-phasestimulus was for the line,sine, and pulse. The cross-modalitymatches, made with each not seenin the responsesof somemid-frequency-sensitive AP and BP stimulustype serving as the standard,showed similar regression effects. fibers,although low-frequency-sensitive AP fibers responded similarly to Thesedata are discussedwith referenceto transitivityand the relation- bothstimuli. Behavioral experiments suggest that bullfrogs can discrimi- shipsamong the data setpredicted from the absolutescaling hypothesis. natebetween in-phase and random-phase versions of theiradvertisement [Work supportedby NIH. ] calls. S96 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S96 MM11. Post-dischargerecovery effects in gerbil auditory-nerveperiod MM14. Auditory-nerve discharge characteristicsthat distinguish and PST histograms. Walter R. Bosch (Institute for Biomedical betweenbasilar papilla neuronsand amphibianpapilla neuronsin the Computing,Washington University, St. Louis,MO 63110) leopardfrog, Rana piplens.Don A. Ronken (Institute for Biomedical Computing,Washington University, St. Louis,MO 63110) Excitation-function(EF) andrecovery-function (RF) parametersof a self-excitingpoint process (SEPP) modelfor auditorynerve discharges The frog'sinner ear hastwo sensoryorgans for auditoryreception. wereestimated [M. I. Miller, J. Acoust.Soc. Am. 77, 1452-1464(1985) ] Oneorgan, the amphibian papilla (ap), hasmany of thecomplexities of for neuraldischarges recorded in gerbilin responseto continuous-tone mammalianears. The otherorgan, the basilarpapilla (bp), is much andtone-burst stimuli. The meandischarge intensity (MDI) of theSEPP simplerand offers special opportunities for comparison to themammalian model(expected value of the PST or periodhistogram) was computed innerhair cell configuration. These anatomical differences make it impor- fromthe EF andRF estimatesby solving an integral equation [ K. Joneset tantto identifythe site of thereceptor cell when recording from individual al., J. Acoust.Soc. Am. 78,90-94 (1985)]. Periodhistograms computed afferentneurons in the main trunk of the auditory nerve.In routineex- fromneural discharges in responseto continuous-tonestimuli were found periments,this can be a dimcultproblem, since the characteristic frequen- to agreealmost exactly with the corresponding MDI overa rangeof stim- cy (cf) of the highest-frequency(ap) neuronsoverlaps the cf of bp neur- ulussound-presSUre levels from 0 to 50 dB abovethreshold, indicating onsin manyspecies. To approachthis problem, a statisticaldata base was that the SEPPmodel accounts well for post-dischargerecovery effects in createdusing responses to tonebursts and clicks,tuning curve character- periodhistograms under these stimulus conditions. PST histograms com- istics,and spontaneousrates. A standardstatistical method, discriminant putedfrom the neural response to tone-burststimulation also agreed quite analysis,produced reliable classification of neuronswithout using cf esti- wellwith thecomputed MDI for theconditions tested. [Work supported matesand offers the possibilityof a fully automateddecision procedure. by NIH grantsNS21592 and RR01379.] [Work supportedby NIH grantsNS21592 and RR01379.] MM15. Cochlearmechanical model stability. John W. Matthews and MM12. Neurophysiologicalobservations on the auditorycortex in the CharlesE. Molnar (Institutefor BiomedicalComputing, Washington FM bat,Eptesicusfuscus. Cynthia F. Moss(Department of Psychology, University,St. Louis, MO 63110) Harvard University, Cambridge, MA 02138), JonathanFritz, MichaelFerragamo, and James A. Simmons (Department of The introductionof nonlineardamping into computationalsimula- Psychologyand Section oœNeurobiology, Brown University, Providence, tions of stable active cochlear mechanical models has introduced instabil- Rl 02912) ity in computationalsimulations of modelsused here. Nonlinear dynami- calsystems theory provides tools that give insight into the possible sources The echolocatingFM bat, Eptesicusfuscus, integrates target range of suchinstability. Analysis of differentialequations for a simplifiedcoch- and shape(range profile) into unifiedimages along a synthetictarget- lear mechanicalmodel with a computerprogram ("AUTO" by Eusebius rangeaxis. The responseproperties contained in single-and multi-unit Doedel) suggeststhat the unstablebehavior o• earlier simulationsis an recordingsfrom the auditory cortex of Eptesicusin relationto the process artifactof the computationsrather than an inherentfeature of the differ- of imageformation revealed in psychophysicaldata wereexamined. A entialequation model. [Work supportedby NIH grantsNS21592 and tonotopicmap wasfound in primaryauditory cortex similar to what has RR01379.] beendescribed previously, but alsoa muchlarger auditory cortical repre- sentationthan hithertoreported. Echo delay-tunedneurons are inter- spersedwith frequency-tunedneurons, showing no apparent topographic organization.The responseprofiles of someecho-delay-tuned and delay- trackingneurons are dependenton the harmonicstructure of the FM MMI6. Perceptual fusion of tonal and noisy sounds.Dik J. Hermes stimuli,as required for representingthe unusualdual time- and frequen- (Institute for PerceptionResearch/IPO, P.O. Box 513, NL 5600 MB cy-domainimages that the bat perceives.Strong suppression effects may Eindhoven, The Netherlands) sharpenperceived target range and be thebasis for encodingthe depth structureof sonartargets in Eptesicus. If stationarynoise is added to a tonal,periodic sound, the noise is, to a largeextent, perceived as a separatesound coming from a sourceother thanthe periodic sound. Perceptual fusion of noisewith a periodicsound requiresspecific conditions. Results will be presentedthat indicatethat the temporalenvelope of the noiseplays an importantrole in thismatter. In experimentsin whichthis envelope had the same peridocity as the tonal MM13. ABER iatenciesare prolongedin rat pupsfed a biotin deficient stimulus,perceptual fusion did occur. Furthermore, the energy content of diet. L. P. Rybak, C. Whitworth, V. Scott, and B. Bhardwaj thenoise within successive periods of theenvelope was made more or less (Departmentof Surgery,SIU Schoolof Medicine,Springfield, IL 62794- constant.These results were obtained in an experimentalsetup in whicha 9230) quantitative measurewas obtainedfor the extent to which the noiseinte- Theeffects ofbiotin deprivation on ABER werestudied in rat pupsfed gratedwith the periodicsound. It will be arguedthat perceptualfusion as a dietthat lackedthis compound. Pregnant rats were either fed a normal reportedhere cannotbe explainedby a peripheralmechanism such as rat diet (controls) or a specialdiet without biotin (Teklad, Harlan adaptation,but mustbe the result of a centralprocess which groups com- SpragueDawley, Madison, WI). Afterweaning, the respective diets were ponentsfrom a wide rangeof frequencybands into one or more sound continuedin pupsfrom each group. ABER wastested under chloral hy- images. drateanesthesia (380 mg/kg) usingclicks ( 100/zs,5/s) andtone bursts (3-msduration, 1.5-ms rise-fall time at 2, 8, 16, and 32 kHz) presented freefield with a driversuspended 8 cm abovethe vertexin the median plane.Responses were obtained using an active lead œrom the vertex with a MMI7. Vowel formant frequenciesproduced with and withoutauditory referenceelectrode inthe nose. Blood levels of biotin were measured using feedback.Emily A. Tobcy and Christy Murchison (Department of a microbiologicalassay. Animals on thebiotin-frec diet showed alopecia CommunicationDisorders, Louisiana State University Medical Center, and dry, scalyskin. There wasno significantdifference in ABER thresh- New Orleans, LA 70112) oldin controlversus experimental animals. However, there was a signifi- cantlylonger brain-stem transmission time (interpeaklatencies from Severalauthors have hypothesized that auditoryfeedback is usedto wavesI to IV) in biotin-deficientanimals. •Ehe lowering of serumbiotin in refinearticulatory gestures in childrendeveloping speech. In order to expcrimentalswas confirmed by microbiologicassay. These findings sug- examinethe roleof auditoryfeedback in children,vowel production was gestthat biotin deficiency in thedeveloping rat adverselyaffects auditory- examinedin 12 childrenranging in agefrom 3-13 yearswho receiveda nerveand/or brain-stemdevelopment. These effects could be mediated by multichannelcochlear implant. Data werecollected 1 yearafter using the a delayedmyelination of thesestructures. [Work supportedby Deafness implant.Speech samples consisted of tenrepetitions of theword, "head," ResearchFoundation. ] spokenunder four conditions:implant on, immediatelyafter the implant S97 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $97 wasturned off, 20 min after nonuse,and immediatelyafter the implant singlepower function representation of puretone intensity-jnd data can wasturned on. Formantfrequencies, fundamental frequencies, and dura- generatea loudnessfunction from a McGill•3oldbergtype model in bet- tion weremeasured. Significant shifts in vowelformant frequencies were ter agreementwith loudnessmeasurements than has heretofore been dem- foundwhen the implant was turned off (p < 0.0001) andreturned to more onstrated.This improvementarises from the presenceof an integration nearly normal valueswhen the implant wasturned on. Two of the youn- constantthat had beenpreviously ignored. The inclusionof the integra- gestsubjects were unable to producestimuli after 20 min of nonusewith tionconstant markedly improves the overallshape of thederived loudness the implant. Data supportthe hypothesisthat auditory feedbackmay function,somewhat more at low then at high intensitylevels. Using the assistin refiningthe articulatorygestures of somespeech sounds. samenear misspower function, the consequencesof settingthe integra- tion constantequal to zero are indicated.These effects are comparedto thoseobtained from the unalteredintensity-jnd data [W. S. Hellman and R. P. Hellman,.I. Acoust.Soc. Am. Suppl.1 82, S25 (1987) ]. Corrections resultingfrom goingbeyond first order in thecalculational procedure are MM18. Extracting stable measures from unstable psychometric discussed.[ Partially supported by the RehabilitationR&D Serviceof the performance.Marjorie R. Leek (Army Audiologyand SpeechCenter, VA.] Walter ReedAMC, Washington,DC 20307-5001),Thomas R. Hanna, and Lynne Marshall (NSMRL, Box 900, SUBASE NLON, Groton, CT 06349) At thelast meeting of the AcousticalSociety of America,a newmeth- od of estimatingslopes of psychomctricfunctions when thresholds were MM21. Increased CMR resulting from lowered levels of masker band. varyingover the course of an experimentwas reported. Computer simula- Marion F. Cohen (Department of Communication Sciences,The tionswere used to calculatethe relationshipbetween the trncslope of the Universityof Connecticut,Storrs, CT 06268) functionand the variability across pairs of trialsin twointerleaved adap- Comodulationmasking release (CMR) for a 1000-Hz, 50-mssinusoi- tive tracks.Further, the variabilityacross all trialsin the two trackswas dal signalwas measured in the usualmanner, except that the overall level shownto indicatethe stabilityof the functionover time. Here, the valida- of the 100-Hz-widemasking noise band was varied from 18to 78 dB SPL tion of that procedurein the measurementof humanperformance is re- in 20-dB steps.The cue band was always78 dB SPL. Presentationwas ported.Listeners were asked to detecta 1000-Hz tonein a narrowband of eithermonotic or dichotic,with the cueband presented to the nonsignal noise,the level of whichwas either fixed or variedsinusoidally from trial ear only. Resultsshow CMR valuesaveraging as high as 7 dB when the to trial. Thismasker variation imposed systematic threshold shifts analo- cueband was centered at 8000Hz, threeoctaves away from the signal, and gousto thosein the simulations.The periodof sinusoidalshift ranged 12dB whenit wastwo octavesaway. CMR wasgreatly reduced when the from 16-128trials per cycle, and the amplitudeof shiftwas either 4 or 8 two noisebands were at equallevels. These results were essentiallyrepli- dB. Psychometricfunctions for individual listenerswere constructed both cateddichotically, except that CMR was somewhatsmaller. These data by traditionalmethods of fittingthe trial-by-trialdata, and by usingthe relate to previousresults showing poorer detectabilityof a noise-band functionalrelationship between slope and variability determined from the signalwhen a synchronousnoise band is introduced at a distantfrequency computersimulations. When the imposedthreshold shift did not occur [M. F. Cohen and E. D. Schubert, .I. Acoust. Soc. Am. 81, 721-723 toorapidly, the newermethod provided more stable slope estimates than (1987); D. McFadden, J. Acoust. Soc. Am. 81, 1519-1527 (1987)]. did the traditionalfitting methods. Under certain conditions the proce- [Work supportedby AFOSR.] dure wasable to identifythe unstableperformance. These results show that this interleavedtracking technique can producemore stable slope estimates,and can alert experimenters to driftingin subjectperformance. MM22. Homophasieand antiphasiemazking as a functionof signal bandwidthand frequency region. John P. Preece,David A. Eddins,and MMI9. Prior auditory stimulation influencesthe precedenceeffect. RichardH. Wilson (Audiology,VA MedicalCenter, Long Beach, CA Ruth Y. Litovsky, Richard L. Freyman, Urea Balakrishnan, and 90822 and Universityof California, Irvine, CA 92664) Rachel K. Clifton (University of Massachusetts,Amherst, MA 01003) Thresholdsfor homophasicand antiphasicmulticomponent tonal Previous studies have demonstrated that echo threshold for clicks is complexeswere measuredas a functionof bandwidth in broadbandnoise elevatedby a precedingclick train. The presentstudy examined the influ- to replicateand to extendthe classicmasking experiment [G. G•issler, enceof a precedingclick train for threevariables presented in 16combina- Acoustiea4, 408-414 (1954) ]. Giisslerreported that maskedthresholds tions: click rate (1-16/s), number of clicks in the train (3-17), and dura- for tonalcomplexes were nearly constant for bandwidthsless than a nomi- tion of train (0.5-8.0 s). Subjectswere testedin an anechoiechamber, nal "critical bandwidth,"beyond which thresholdsincreased 3 dB/oct. with two speakerspositioned at 45 degleft andright of midline.Leading Three experimentswere conducted in broadbandnoise employing (a) clickswere alwayspresented from the left speaker,while laggingclicks homophasicsignals similar to G•issler's( 1 to 20 sinusoldsspaced 20 Hz werepresented from the right speakerwith delaysranging between 2-14 apart extendingdownward from 1100Hz); (b) homophasicand antipha- ms.Each trial consisted ofa clicktrain followed by the test click. Subjects sicsignals composed of 1 to 23 sinusoidscentered at 500, 1000,and 2000 presseda buttonwhen they heardthe laggingtest click on the right. Sub- Hz; and (c) homophasicand antiphasicsignals centered at 500 Hz with jects'echo threshold was defined as the clickdelay at whichthe echowas bandwidthsto 800 Hz (71 sinusolds).The slopeof the threshold/signal reportedon 50% of trials. Relativeto an isolatedtest click, thresholds bandwidthfunction beyond 100 Hz was < 3 dB/oet at each centerfre- wereelevated approximately 3 to 6 mswhen preceded by a clicktrain. The quency,ranging from 0.3 dB/oct (2000 Hz) to 1.9 dB/oct ( 1000Hz). mostcritical variable influencing magnitude of shiftin echothreshold was The dataindicate that the integrationof signalenergy continued beyond the numberof clicksin the precedingtrain. For a constantnumber of theso-called "critical band" estimates reported by Giissler. There were no clicks, duration of the train had little effect, whereas rate sometimes inter- substantialdifferences between the slopesof the homophasicand antipha- actedwith number.[Work supportedby NSF.] sicconditions at 500, 1000,and 2000 Hz. [Work supportedby VA Medi- cal ResearchService. ] MM20. Is a powerfunction description of intensity-jnddata compatible with the loudnessfunction? William S. Hellman (Department of MM23. Psyehometricfunctions for unattendedtones. Huanping Dai, Physics,Boston University, Boston, MA 02215) and Rhona P. Hellman Soren Buus, and Bertram Scharf (Northeastern University, Boston, (Department of Psychology,Northeastern University, Boston, MA MA 02115) 02115) Listeners detect attended tones better than unattended tones. To de- A sequenceof approximationfor calculatingloudness functions from terminehow muchbetter, percentage correct detection was measured as a the intensity-jndfunction for a l-kHz toneis performed. It isshown that a functionof signallevel for tonesat 24 unattendedfrequencies, all more S98 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S98 than a criticalband from the attendedor primaryfrequency of 1000Hz. A physiologicallybased auditory model [O. Ghitza,J. Phon.16{ 1 ), The primarywas presented as a cueat the beginningof everytrial 109-123(1988)] wastested against human psychophysical data. The 2AFC) andalso as the signal on 67% of the trials.On therest of thetrials, modelconsists of a setof cochlearfilters inferred from frequency selectiv- the signalcould be anyone of 24 differentprobe frequencies. All signals ity of primaryauditory fibers in cat,rate-level processing, and synchrony werepresented against a wide-bandnoise at 62 dB.Nearly 6030 trials by extraction.Isoioudness contours for puretones are calculatedfrom the threeexperienced subjects yielded a meanpsychometric function for the modeland are comparedto thosemeasured in humanby Fletcherand probesthat wasparallel to that for the primary,but shiftedto the right, Munson(1933) andby Robinsonand Dadson { 1956).The model is also showingpoorer detection of theprobes. Closer analysis revealed that the usedto simulateScharf's [J. Acoust.Am. 31, 365-370 (1959)] critical- 14probes at frequenciesabove 1000 Hz were"attenuated" 2 to 3 dB,and bandmeasurements using complex sounds of fourcomponents. In addi- the 10 probesbelow were attenuated 4 to 6 dB. Theseand earlier results tion, both calculatedand measuredcritical-band resultsare plotted as a suggestthat probe frequencies may be effectively attenuated 6 dB or more. functionof cochlearplace using formulas for humanand cat cochlear [Work supportedby NIH.] mapsin Greenwood'srecent paper (submitted to J. Acoust.Soc. Am. ). Thecomparison of calculatedand measured data suggests possible im- provementsof the model.a• Work done as a consultantto AT&T Bell Laboratories,Acoustics Research Department. MM24. Ftmefional involvement of auditory cortex in fine frequency discrimination of biosonar signals in the bat. H. Riquimaroux, S. J. Gaioni, and N. Suga (Department of Biology, Washington University,St. Louis, MO 63130) The auditorycortex of the mustachedbat, Pteronotusparnellii,has a highlyspecialized region, the DSCF (Doppler-shiftconstant frequency) area, which overrepresentsthe frequenciesof Doppler-shiftedechoes MM26. Effectsof auditory segregationof a harmoniccomponent on (about 61 kHz), and, hence,may play a role in the fine frequencydis- discriminationof dintic intensityand of interauraltime difference.W. criminationnecessary for detailedvelocity measurements ( Suga, 1984). S. Woods and H. S. Colburn (Boston University, 48 Cummington This functionof the DSCF area wasbehaviorally tested using reversible Street, Boston, MA 02215) ablationwith muscimol,a potentGABA agonist(Hilosaka et al., ! 985). Researchershave demonstratedthat the just-noticeabledifference Batswere trained on a discriminatedshock avoidance task requiring a leg (jnd) for interauraltime difference (ITD) of a tonaltarget increases in flexionresponse. The stimuli were trains of artificial pulse-echopairs the presenceof tonal maskers,even with target/maskerseparations (tone bursts).For S +, the pulseand echowere the samefrequency (e.g., greaterthan a criticalband [ Buelland Hatter, J. Acoust. Soc. Am. Suppl. 61.0 kHz), while,for S --, they weredifferent frequencies (e.g., 61.0 and 1 83, S!22 ( 1988);Stellmack et al., J. Acoust.Soc. Am. Suppl.I 85, S83 61.1kHz). Followingbaseline testing, 0.1-0.2/x[of muscimol( (1989) ]. In mostcases, subjects report the perceptionof one "sound saline) wasbilaterally applied to the DSCF area. The bats failed on pre- object,"whose subjective position is affected by the ITD at eachfrequency viouslysuccessful discriminations of smallfrequency differences (e.g., 50 of the stimulus.It hasalso been noted that asynchronizingthe onset times Hz), but succeededon discriminationsof large frequencydifferences of componentsofa stimulusyields more than one "sound object." This (e.g., 2 kHz). Discriminationperformance returned to baselinelevels experimentinvestigates whether this parsing can lead to unmaskingof a within 24 h. Thesedata indicate that the DSCF area is necessaryfor fine component,and whether a general model of parsing can be used to predict frequencydiscriminations involving biosonar signals. [Work supported observedeffects. A 41, 2AFC adaptiveprocedure was usedto measure by AFOSR Grant No. 8743250.] dinticintensity and ITD discriminationjnd's for a 100-ms,600-Hz target, bothin isolationand when masked by a two-tonemasker comprising 400- and 800-Hz tones,under varying conditions of relativeonset time of MM25. Evaluation of a physiologicallybased auditory model using maskerand target. Although generally lower jnd's were obtained for the psyehophysicaldata. Patricia S. Jeng*) (Centerfor Researchin Speech asynchronouscase, the factthat thedifference was small relative to the and HearingSciences, City Universityof New York, New York, NY variabilityin theadaptively measured jnd's prompted a fixed-increment 10036) and JosephL. Hall (AcousticsResearch Department, AT&T measurementof sensitivityunder the same conditions. [Work supported Bell Laboratories,Murray Hill, NJ 07974) by NINCDS. [ THURSDAY AFTERNOON, 30 NOVEMBER 1989 ST. LOUIS BALLROOM A, 1:00 TO 4:39 P.M. SessionNN. SpeechCommunication VIII: SpeechPerception David B. Pisoni, Chairman SpeechResearch Laboratory, Department of Psychology,Indiana Unioersity,Bloomington, Indiana 47405 Contributed Papers 1:00 Previouswork hasshown that the segmentalrepresentation used in recognizingspoken words does not correspondto the traditionalabstract NNI. Convergingevidence on the nature of the segmentalrepresentation phoneme[Gagnon and Sawusch { 1989)]. However,the evidence did not underlying spoken word recognition. Deborah A. Gagnon and James unequivocallydistinguish between two alternative proposals--allo- R. Sawusch(Department of Psychology,SUNY at Buffalo,Buffalo, NY phonesand position-specificphoneroes. In the previousstudy, subjects 14260) heardnatural CVC prime-targetpairs in variousdegrees of phoneticover- S99 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall 1989 118thMeeting: Acoustical Society of America S99 lap and namedthe seconditem (the target) as quicklyas possible.The /1/were used.The first three formants of each vowel varied in seven equal presentstudy adopted the sameapproach but utilized a phonememoni- logarithmicsteps from/i/to/I/as theduration varied from short to long toringtask in which subjectsresponded whenever the targetbegan with a in fivesteps (50, 80, 120, 190,and 300 ms). All of the vowelstimuli were designatedphoneme. The patternof RT results,while differentfrom that presentedto subjectsunder high- and low-attention conditions. Attention obtainedin the namingtask, again supported a rejectionof the phoneme wasmanipulated by requiring that subjects perform a nonspeechdistrac- andshed further light on thenature of therepresentation. Similar patterns tor taskwhile simultaneously performing a speechidentification task, or across word and nonword blocks and across trials in which voice was the by requiringthat subjects perform a speechidentification task only. Pho- sameversus different within a prime-targetpair were found with both neticidentification of the vowelstimuli was found to varywith the atten- tasks.A comparisonwill be madeof the two task typesin termsof the tion condition.When subjectsperformed the distractortask and the resultsgenerated and the implicationsfor their use in studyingon-line speechtask simultaneously, duration became a moreimportant cue to spokenword recognition.[ Work supportedby NINeDS. ] phoneticidentity, whereas the effect of formantfrequency was reduced. A quantitativemodel was developed to characterizehow the integrationof informationfrom the two kindsof cues,formant structure and duration, changedwith attentionlevel. 1:12 NN2. The role of phonologicalpermissibility in the phoneticcoding of speeeh.CarolA. WannemacherandJamesR.Sawusch (Department of 1:48 Psychology,Park Hall, SUNY/Buffalo,Buffalo, NY 14260) NN$. The intellig•bilityof speechdirected to childrenand adults.Judith Previousresearch has shown that phonological permissibility of a con- C. Goodman, Howard C. Nusbaum, Lisa Lee, and Kevin Broihier sonantsequence can affectthe identificationof an ambiguousphoneme (Committeeon Cognitionand Communication,University of Chicago, [Massaroand Cohen (1983)]. The presentexperiment used both reac- 5848 S. UniversityAvenue, Chicago, IL 60637) tiontime and identification measures to furtherexplore this finding. With naturalspeech, little or no effectof phonologiealpermissibility of a frica- Speechdirected to very youngchildren differs markedly from speech tive-liquidsequence was found upon speed to classifythe liquid.In the directedto adultsin vocabulary,syntax, and intonation.Are therediffer- secondexperiment, a syntheticliquid serieswas constructedwith end- encesin segmentalintelligibility as well?Speech was recordedfrom six pointsvarying in permissibility(depending on thefricative precursor). A mothersaddressing their 2 yearolds, and, in anothersession, addressing speededclassification task was run and data were partitioned according to anadult, under comparable conditions. Open and closed-class words were speedof response(cf. Fox, 1984). The resultsof thisexperiment, com- excisedfrom both typesof speechand were presented,in isolation,to binedwith the earlierfindings, illustrate the relativeroles of both acous- adult listenersfor identification. Children may not understandclosed- tic-phoneticand phonological information in phoneticcoding. The data classwords, so mothers may not articulate them clearly. Excised open- will be discussedin termsof the data- and knowledge-driveninfluences classwords expressed either new or previouslygiven information. Where- uponphonetic coding. {Work supportedby NINCDS.] as repeatedinformation is articulatedless clearly in speechto adults, repeatedinformation in speechto childrenmay be intendedto aid com- prehension.Thus the given/newdistinction may affectthe intelligibility of speechto adultsand to childrendifferently. Preliminary results indi- catethat child-directedspeech may be less intelligible than adult-directed speech,particulary for closed-classwords: Expectations about children's 1:24 linguistic knowledge apparently affect articulation of different word classes.[Work supportedby NIH and a BiomedicalResearch Support NN3. Attention to phonetic context across word Grant. ] boundaries.Jenny DeGroot and Howard C. Nusbaum(Department of Psychology,University of Chicago,5848 S. UniversityAvenue, Chicago, IL 60637) 2:00 Variabilityin phoneticcontext affects phoneme perception. For exam- ple, subjectsare slowerto identifya targetconsonant when an adjacent NN6. Cues to perceptual normalization of talker differences.Todd contextphoneme varies independently, compared to whencontext pho- M. Morin and Howard C. Nusbaum{Department of Psychology,The nemesare constantacross trials [C. C. Wood and R. S. Day, Percept. Universityof Chicago,5848 S. UniversityAvenue, Chicago, IL 60637) Psychophys.17, 346-350 (1975) ]. Further, this is true whetherof not a syllableboundary intervenes between the targetand the varyingcontext, To recognizevowels produced by a specifictalker, a listenermust suggestingthat subjectsattend to phoneticcontext regardless of syllable know somethingabout the talker's vowel space.How do listenerslearn structure[J. DeGroot and H. C. Nusbaum,J. Acoust.Soc. Am. Suppl. 1 abouta talkers'vowel space? How differentmust two talkers'vowels be to 85, S 123 (1989) ]. Do listenerssimilarly attend to phoneticcontext across constitute different vowel spaces?To investigatethese questions, listeners a word boundarywhen recognizinga phoneme?In the presentexperi- monitoredsequences of targetvowels for a specifiedtarget. In one condi- ments, subjectsidentified a target phoneme,while an adjacentcontext tion, a singletalker producedthe vowelsfor eachtrial. In a secondcondi- phonemeeither varied orthogonally or washeld constant. In onestimulus tion, a mix of four talkersproduced the vowelsfor each trial. Previous set, the varyingcontext occurred in the sameword as the target;in the researchdemonstrated that vowelrecognition is accurate when talkers are other stimulusset, the vrying contextoccurred in an adjacentword. Re- mixed,but it requiresmore attention. The presentstudy compared recog- sponsetimes were measuredto investigatewhether the varyingcontext nitionfor whisperedand voicedvowels in theseconditions to eliminatethe slowstarget recognition across a wordboundary as it doeswithin a word. useofF0 asa cuefor talkernormalization. Another experiment examined The resultsprovide information about how listenersdistribute their atten- the effectsof talker differenceson normalization. Performancefor a pair tionacross the speechsignal, and about the perceptualfunction of differ- of talkerswith similarvowel spaces was compared to performancefor a ent linguisticunits. [Researchsupported by NIH.] pair of talkerswith more disparatevowels. The resultsof thesestudies indicatethe importanceof fundamentalfrequency in vowel recognition acrosstalkers and that the amountof effortrequired for normalizationof talker differencesdepends on the similarityof the differenttalkers' vowel spaces.[Work supportedby NIH.] 1:36 NN4. The effects of attention on the phonetic integration of acoustic information. Jennifer L. Eberhardt and Peter C. Gordon (Department 2:12 of Psychology,Harvard University,Cambridge, MA 01238) NN7. On the perceptual representation of vowel categories. The presentstudy investigated the effectsof attentionon phonetic Keith Johnson{Speech Research Laboratory, Psychology Department, processing.Thirty-five synthetic steady-state vowels varying from/i/to IndianaUniversity, Bloomington, IN 47405) S100 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S100 Subjects'behavior in a vowelperception task was compared with the 2:36 predictionsof twodifferent types of vowelperception models. Both mod- elsuse auditory spectra generated by a modelof theperipheral auditory NN9. Comprehensionof naturaland synthetic speech. James V. Ralston, system[Bladon and Lindblom,J. Acoust.Soc. Am. 69, 1414-1422 John W. Mullennix, Scott E. Lively, Beth G. Greene, and David ( 1981) ]. In thefirst model, it isassumed that the spectral information in B. Pisoni(Speech Research Laboratory, Department of Psychology, vowelsis represented asa wholespectrum. The spectral characteristics of IndianaUniversity, Bloomington, IN 47405) a vowelcategory are modeled as a spectraltemplate against which incom- Previousstudies comparing the comprehensionof natural and syn- ingvowel spectra are compared. In thesecond model, the spectral infor- theticspeech passages have produced conflicting results. All of thesestud- mationin vowelsis characterizedby thefrequency locations of spectral iesused successive measurement techniques in whichsubjects' compre- peaks.Here, the spectralcomponent of a vowelcategory is takento be hensionwas assessedafter the presentationof a passage.However, informationconcerning the frequencies of spectralpeaks. Two sourcesof comparativestudies have found that successivemethods are lesssensitive voweltemplates for the modelswere also compared. In one case,tem- than simultaneous "on-line" measures. Successive measures are also plateswere derived from the subject'sown productionsof the vowelsin knownto be influenced by memory factors. Subjects in thepresent experi- question.In the other, templateswere derived from vowelssynthesized mentmonitored for word targets while they listened to shortpassages and withaverage formant values reported by Petersonand Barney [J. Acoust. thenverified statements after each passage. Both monitoring and verifica- Soc.Am. 24, 175-184 (1952) ]. Subjectswere instructed to find the best tion performancefor passagesof syntheticspeech were depressed relative exemplarsof eachof 11different vowel categories from an array synthetic to passagesof naturalspeech. There wasa significantinteraction between steady-statevowel tokensby adjustingthe F 1 and F2 of the tokensuntil voiceand text difficultyin the monitoringlatency data, suggestingthat the bestexemplar was located. In all cases,it wasfound that the whole bothfactors affect mechanisms drawing from the same limited processing spectrumapproach provided the best fit to thesubjects' perceptual judg- resources.Finally, therewas a significantinteraction between voice and ments.It wasalso found that model predictions based on thePeterson and sentencetype in the verificationtask. Subjectslistening to passagesof Barneyvowels provided a betterfit to theperceptual data than did predic- syntheticspeech had relatively poor memory for propositionalinforma- tionsusing the subject'sown productions as the basisfor spectraltem- tioncompared to surface(word) information.Taken together, the results plates.This finding suggests that perceptual vowel categories are not sub- indicatethat comprehensionof syntheticspeech is poorer and slower than ject to theidiosyncrasies of the individual'sown speech production, but natural speech.Memory for propositionalinformation extracted from ratherare a productof thelistener's range of theperceptual experience. passagesof syntheticspeech is particularlypoor. [Work supportedby NSF.] 2:24 2:48 NNS. The intelligibility of LPC-vocodedwords and sentencespresented to native and non-nativespeakers of English.Molly Mack (Division of NN10. Perceptionof Hindi retroflex versusdental stopsby monolingual Englishas an InternationalLanguage, 3070 Foreign Languages Building, speakersof American English. Linda Polka (Speech Perception 707 S. Mathews,University of Illinois, Urbana, IL 61801 and the MIT Laboratory, Department of CommunicationSciences and Disorders, Lincoln Laboratory,Lexington, MA 02173) Universityof SouthFlorida, Tampa, FL 33612) The presentstudy is an extensionof previouswork [M. Mack, J. Cross-languagestudies have shownthat foreignconsonant contrasts Ncurolinguist.3, 293-316 ( 1988); M. Mack., J. Acoust.Soc. Am. Suppl. vary in the degreeof perceptualdifficulty that they presentadult non- 1 81, S4 (1987) ] designedto assessthe intelligibilityof naturaland com- nativelisteners. Phonetalc, phonetic, and acousticfactors have been con- puter-generatedspeech presented to native and non-nativespeakers of sideredimportant in accountingfor this variability. Thesefactors were English.In this study,subjects were presented with threetests--the Diag- examinedby comparingEnglish listeners' perception of the Hindi retro- nosticRhyme Test (DRT), a meaningfulsentences test, and a semantical- flex versusdental place-of-articulationcontrast in four differentvoicing ly anomaloussentences test. Stimuli were presentedin two conditions-- contexts: prevoiced, voicelessaspirated, voicelessunaspirated, and naturaland LPC (2.4 kbps)vocoded. Subjects were 20 nativespeakers of breathyvoiced. Differences in the perceptualdifficulty of the four Hindi English and 20 native speakersof German, fluent in English. Results contrastswere predictedbased on: ( 1) phonetalcstatus (the functional revealedthat the non-nativesperformed significantly worse than the na- statusof thecon trast in listeners'native phonology ), ( 2 ) phoneticfamil- tiveson all but the natural-speechDRT; word frequencyaffected accura- iarity (as allophonesor freevariants), ( 3 ) differencesin acousticsalience cy more for the non-nativesthan natives;and in the vocodedcondition, related to voicing,and (4) assimilationstrategies. Differences in perfor- the non-nativesappeared to exhibitfatigue effects, unlike the natives,who mancein a categorialAX discriminationtask were ordered from mostto exhibitedperceptual learning. Moreover, the effectsof testtype and con- leasterrors: prevoiced, voiceless aspirated, breathy-voiced, and voiceless ditionupon non-native response accuracy were nonadditive. Implications unaspirated.Perceptual differences were correlatedwith both acoustic for modelsof non-nativespeech processing and the useof codedspeech salienceof placecues and subjects'descriptions of their assimilationstra- systemsby non-nativelisteners are considered. tegies.[Work supportedby NICHD and NINCDS.] 3:00-3:15 Break (unmodified)or truncatedtokens of naturallyproduced breathy-voiced 3:15 [•1ha] vs [d_ ha]. Interstimulus intervals (ISI) wereeither short ( 550 ms) NNI1. Effects of auditory and phonetictraining on Americans' or long (1500ms). Conditionsfavoring auditory-processing strategies discriminationof Hindi retroflex-dentalcontrasts. Winifred Strange, (shortISI andtruncated syllables) yielded the most improvement during Linda Polka,and Manuela C. Aguilar(Speech Perception Laboratory, training.However, pre-test to post-testimprovement in "phonetic-level" Departmentof CommunicationSciences and Disorders,University of discriminationof the full syllableswas not significantlydifferent across South Florida, Tampa, FL 33620-8100) thefour training conditions, and there was no transfer to discriminationof thecontrast invoiceless aspirated [t h a] -- [ffa] orprevoiced [•la]-[_da] Previous research has shown that the Hindi retroflex-dental contrast stops.Large individual differences Pound ineach training condition sug- amongstop consonants is not easilydifferentiated by English.speakers gestthat subjects' strategies may be moreimportant than stimulus and evenafter some training. In thepresent study, subjects were given 4 days taskvariables in predictingsuccess in perceptualdifferentiation of non- of trainin•g(768 trials) in a cate_gorialAX discriminationtask on full nativecontrasts. [Supported by NINCDS.] S101 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S101 3:27 4:03 NNI2. Training methods for the facilitation of Japanese students' NN15. Spectral slopeas a cue for the perceptionof breathy and non- perception of American English /r/ and /1/. Salvatore Miranda breathystops in Shanghalnase.Nianqi Ren and IgnatiusG. Mattingly (Departmentof Linguistics,University of Connecticut,U-145, Storrs, (Haskins Laboratories,New Haven, CT 06511 and Department of CT 06269-1145), [AV: Melva Underbakke, Winifred Strange, and Linguistics,University of Connecticut,Storrs, CT 06269-1154) Theodore Micceri (Speech Perception Laboratory, Department of In Shanghainese,breathy excitation following release distinguishes CommunicationSciences and Disorders,BEH 255, Universityof South category 2 (so-called"voiced unaspirated") stopsfrom category 1 Florida,Tampa, FL 33620-8100) ("voicelessunaspirated") stops. The mostobvious acoustic correlate of Thisstudy investigated the effects of two typesof discriminationtrain- bresthinessis relativespectral slope: During the first 50-100 ms after ingon theperception and production of/r/and/1/by Japanesestudents release,H I-H2 isgreater for category2 thanfor category1 in bothword- of English.Training stimuli cansisted of syntheticsyllables drawn from initial and word-medial(morpheme-initial) stops (Ren, unpublished). three differentten-step/r/-/1/series: rock-lock, rake-lake, and rook- To investigatethe perceptionof breathiness,a series of syntheticdisylla- look.The prototypegroup received discrimination training using only the bleswas prepared in whichthe spectral slope during the 100ms following endpointstimuli from each series (the clearest instances of/r/and/1/). the releaseof the word-medialstop was increased step by stepincreasing The gradientgroup received discrimination training that beganwith the the valueof the "openquotient" in the computationof the glottalwave- endpointsand proceeded gradually to stimulinear the (English)category form--a techniquepreviously used for the synthesis of breathyvowels [C. boundaries.Control subjects received discrimination training on natural- Bickley,Work. PapersMIT SpeechCommun. Lab. 1, 71-81 (1982) ]. ly producedtokens of/b/vs/v/ (anothercontrast that Japanesefind Other potentialcues (closure duration, pulsing during closure, tone con- difficult).Comparisons of pre-testand post-test scores showed improve- tour) were neutralized.Ten Shanghainesespeakers labeled the stimuli mentfor bothprototype and gradient groups in identificationof synthetic with greaterspectral slope as category 2 and thosewith smallerspectral speechseries and in production,but not in perceptionof naturallypro- slopeas category 1. [Work supportedby NICHD Grant 01994.] ducedminimal pairs.Improvement for the controlgroup was restricted to the perceptionof the synthetic/r/-/1/series. No differenceswere found in the overalleffectiveness of prototype versus gradient training. [Sup- portedby NINCDS.] 4:15 NNI6. The separation of two voicelike signals. P. G. Vaidya (Departmentof Mechanicaland Materials Engineering,Washington 3:39 StateUniversity, Pullman, WA 99164) NNI3. Perceptual characteristicsof English syllable-initial /r,I/ for If two personsare talking or singingat the sametime, human beings Japanese listeners. Reiko Yamada, Yoh'ichi Tohkura, and canselectively listen to onlyone of them,even if theyare producing notes Noriko Kobayashi (ATR Auditory and Visual PerceptionResearch thatare in unison.This paper presents an attemptto simulatethis task by Laboratories,Inuidani, Seika-cho, Kyoto 619-02,Japan) a computer.An elementarymodel of the voicesof the two speakershas This study investigatesthe perceptualcharacteristics of American beenused. Using this model, two voicelikesignals are generatedand English/r,l/for Japaneselisteners using synthesized stimuli. Five major mixed. The resultingsignal is analyzedfor the coherenceof the trans- findingsare obtained. ( 1) The Japaneselisteners identify the stimuliusing phase.Trans-phase isthe phase difference between the components at two a variety of acousticcues, and their responsepatterns are stronglyin- differentfrequencies when one of themhas been sent through a nonlinear fluencedby acousticfeatures of the stimuli. In contrast,the American filter.If the trans-phaseshows coherence, the signalsare said to betrans- listenerscan identify/r/and/1/as long asa primarycue remains, even spectrallycoherent. A justificationfor the assumptionof suchcoherence underthe conditionwhere some ofthe acousticcues are missing.(2) As in voicelikesignals has beenpresented. The data for the transspectral the Japaneselisteners tend to perceivesome stimuli as/w/more than coherencehave been sent through a processorthat yieldsan estimateof Americanlisteners do, perceptionexperiments with/w/as well as/r/ the individualsound separately. and/1/for a choiceof identificationbetter clarify the perceptionmode of the Japaneselisteners. (3) A positiverelationship between the identifica- tion abilityof the natural/r,l,w/spoken by nativeAmericans and that of thesynthesized/r,l,w/is found for theJapanese listeners. (4) Contextual 4:27 effectsin wordsare very strongfor the Japaneselisteners when trying to identify/r/and/1/. (5) TheJapanese listeners who have lived in English- NN17. The role of auditory object formation in sentenceperception. speakingcountries before a certainage are able to identify/r/and/1/ Thomas D. Carrell (Communication Sciences and Disorders, as well as native Americans. NorthwesternUniversity, Evanston, IL 60208-3570) Previouswork has demonstratedthat amplitude-modulatedtime- varyingsinusoidal (AM-TVS) replicasof naturalsentences were more 3:51 intelligiblethan simple time-varying sinusoidal (TVS) sentences[T. D. NNI4. Perceptual categorizationof synthesized/R-W/ continua in Carrell,J. Acoust.Soc. Am. Suppl.I 84, S158(1988)]. The goalof the normalpreschool children. Laurie Fitzgeraldand Elzbieta B. Slawinski experimentsreported here was to determinethe causeof this increased (Departmentof Psychology,The Universityof Calgary,Calgary, Alberta intelligibility.One potential explanation was based on the factthat AM- TVS sentenceswere rated more natural (i.e., more humanlike) than TVS T3A IN4, Canada) sentences.This increasednaturalness might have increased the probabili- The purposeof the presentstudy was to examinethe developmentof ty that speech-specificanalysis was brought to bearin decodingthe in- categorical perception of phoneme boundaries. A seven-step adult comingmessage. The secondpotential explanation was based on the idea /R-W/continua wassynthesized via the Klatt cascade/parallelsoftware that the comodulationof the three componenttones of a TVS sentence program.The acoustictokens varied according to the second(F2)- and mightcause them to begrouped together as an auditoryobject for further third ( F 3 )- formantonset frequencies and second ( F 2 )- and third (F 3 )- processing.It has beensuggested that "...amplitudemodulation helps formanttransition rates. Percentage of correctresponses on an identifica- form auditoryobjects from complexsound fields" [W. A. Yost and S. tiontask was computed to yielda measureof phonereitboundary location Sheft,J. Acoust.Soc. Am. 85, 848-857 (1989) ]. Evidencewill be present- for adultsand childrenof 3, 4, and 5 yearsof age.The phonemicboundar- edin favorof the secondexplanation. Specifically, the increasein intelligi- iesfell betweenstimuli 3 and 4, at stimulus4 and 5 for 3-, 4-, and 5-year- bility will be shownto berelated to the comodulationof the threecompo- old childrenand adults, respectively. The findingthat therewas a shifting nent tonesand possiblyto the mechanismunderlying comodulation andincreasing steepness in the phonemic boundaries as a functionof age is maskingrelease (CMR). Thesefindings suggest that thecomodulation of supportiveof previousresearch and moreover, of the theorythat a child's the threetones comprising a TVS sentencecreate an auditoryobject and phonologicalsystem is not inherentlydifferent than that of an adult but that auditoryobject formation is importantin fluent speechunderstand- ratheris just a simpleror lessprecise version of the maturesystem. ing. $102 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S102 THURSDAY AFTERNOON,30 NOVEMBER1989 DIRECTORS'ROW 26, 1:30TO 2:45P.M. Meeting of AccreditedStandards Committee S1 on Acoustics to be heldjointly with TechnicalAdvisory Group Meeting for ISO/TC 43 Acoustics,and IEC/TC 29 Electroacoustics D. L. Johnson, Chairman SI Larson-DavisLaboratories, 1681 West820 North, Provo, Utah 84061 StandardsCommittee Sl on Acoustics. Working group chairs will report on their progress inthe preparation of standardson methodsof measurementand testing, and terminology, in physicalacoustics, electroacoustics, sonics,ultrasonics, and underwater sound. Work in progressincludes measurement of noise sources, noise dosimeters,integrating sound-level meters, and revision and extension of soundlevel meter specifications. Opendiscussion of committeereports is encouraged. The internationalactivities in ISO/TC 43 Acoustics,and IEC/TC 29 Electroacoustics,will alsobe discussed. TheChairs of the respective U.S. Technical Advisory Groups for ISO/TC 43 ( H. E. yonGierke), and IEC/TC 29 (V. Nedzelnitsky),will reporton currentactivities of theseTechnical Committees. ISO/TC 43 andIEC/ TC 29 will nextmeet in Copenhagenduring April 1990. THURSDAY AFTERNOON, 30 NOVEMBER 1989 ST. LOUIS BALLROOM H, 2:00 TO 4:10 P.M. Session OO. Underwater Acoustics VI: Arctic Ocean Acoustics Michael J. Buckingham,Chairman MissionManagement Department, Royal Aerospace Establishment, Farnborough, Hampshire GU14 6TD, England Chairman's Introduction--2:00 !noited Papers 2:05 OO1.Physical bases for Arctic ambient noise. Ira Dyer(Department of OceanEngineering, MIT, 5-212, Cambridge,MA 02139) The fundamentalsof noisegeneration by ice fractureprocesses such as occurin the Arctic Oceanare reviewed.Noise events, presumably associated with individualice fractures, are measured and studied in terms oftheir source strengths, characteristic frequencies, event signatures, event densities, and ice morphologies. A usefullynarrow set of possible mechanisms seems to be emerging as a plausibleexplanation for a widevariety of cases.[Work supportedby ONR ArcticAcoustics Program. ] 2:25 002. Experimentalmeasurements ofnormal incidence reflection/scattering from undeformed natural sea ice. R. Francois,K. Williams,T. Wen, and P. Mourad (Applied PhysicsLaboratory, College of Oceanand FisherySciences, University of Washington,Seattle, WA 98195) Usingnarrow-band pulses, experimental measurements ofnormal incidence reflection/scattering from the ice/waterinterface of naturallygrowing first year sea ice have been made. The results cover the frequency rangefrom 20-300 kHz. A sufficientnumber of realizationshave been obtained, using different experimental S103 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall1989 118thMeeting: Acoustical Society of America S103 techniques,that generalizationof the resultscan be made.The techniquesinvolve monostatic experiments, ensonifyingseveral different diameters of cylindricalice surfaces,and undeformedareas of ice, usingtrans- ducerswith a varietyof directivitiesincluding omnidirectional. The growing, or skeletal,layer of iceresults in a gradedporous, rough surface that can be treated using Blot theory to obtainacoustic velocities, a continuous velocityprofile for reflection,and Kirchhofftheory for scattering.The resultsindicate that the mean amplitude reflectioncoefficient follows the approximate equation, R• = 0.40 - 0.157log (f) ( f in kHz), in thetransi- tion regionfrom about2-200 kHz. At lowerfrequencies, the reflectioncoefficient is determinedby the bulk velocityof the ice (the velocitygradient in the skeletallayer being unimportant); at high frequencies,the surfaceroughness and ice/water surfaceimpedance mismatch may both contributeto the meanreflection coefficientobserved. [Work supportedby Office of Navy Technologywith technicalmanagement by NORDA. ] 2:45 003. Acousticscattering by laboratoryscale modeled sea ice. Orest I. Diachok(Naval ResearchLaboratory, Washington,DC 20375) and RobertM. Kille (AmericanUniversity, Washington, DC 20016) Laboratoryscale-modeled experiments were designed to providehighly controlled data for evaluationof existingand guidancefor new theoriesof acousticscattering from Arctic seaice. Resultsfrom previously reportedmeasurements of scattering loss from parallel and randomly oriented sea ice ridges situated on a plate betweenair and water ( McClanahanand Diachok) will be reviewedand newresults of scattering(particularly backscattering)versus angle at both high and low frequencieswill be presented.Implications for acoustic remotesensing of geometricalproperties of seaice at greatranges in the Arctic will bediscussed. 3:05 004. Low-frequencyscatter from a thin rough elastic plate: Theory and experiment. Ivan Tolstoy (Knockvennie,Castle Douglas DG7 3PA, SW Scotland) The generalizedsmoothed boundary condition theory, based upon an original idea of Biot's[ M. A. Blot,J. Acoust.Soc. Am. 44, 1616-1622(1968) ], and developedafter 1979by Tolstoy,offers a flexibleand straight- forwardtechnique for obtainingsimple, formal solutions for low-frequencysound scatter from distributions of scatterersof arbitraryshapes and propertiesof any type at liquid/liquid or liquid/solidinterface. In special casessuch as a perfectlyhard wall with perfectlyhard bossesor cylindricalcorrugations, the boundary condi- tion enablesone to recoverwell-known solutions. After a brief reviewof prior successfulapplications to the interpretationof rough boundarymodel experiments,it is shownhow the methodcan be appliedto the reflectivityand phaseproblems for a planewave incidentupon a thin rough plate satisfyingthe Germain equationon the surfaceof a liquid.The predictionsof thetheory are found to agreewell with recentlyreported modelwork IT. P. McClanahanand O. I. Diachok, J. Acoust.Soc. Am. Suppl. I 84, S218 (1988)]. [Work supportedby NRL and ONR.] Cont•butedPape• 3:25 3:40 005. Measurements of under ice broadband coherence. Roger 006. Measurementsof low-frequencyambient noise in the Greenland W. Meredith ( Naval Ocean Researchand DevelopmentActivity, Arctic Sea;Michael J. Buckingham(Mission Management Department, Royal AcousticsBranch, Stennis Space Center, MS 39529) Aerospace Establishment,Farnborough, Hampshire GUI4 6TD, England) TheArctic is a uniqueregion acoustically, dominated by theunder ice Sonobuoymeasurements of omnidirectionalambient noise spectra canopy.The ice variesseasonally in coverage,thickness, and roughness havebeen made at severalpositions between the DenmarkStrait and the causingvarying degrees of acousticspatial and temporal variability. Hori- Fram Strait off the eastcoast of Greenland.Results from linesof buoys zontal coherencefrom explosiveacoustic sources at a depth of 300 m placednormal to the iceedge indicate that, below200 Hz, the noiselevel is under a relativelysmooth Arctic ice pack is examinedin the frequency essentiallyuniform across the edge,with no peakobserved in the levelat range 100-1500 Hz over source-receiverranges from 3-50 km. Coher- the interfacebetween the ice coverand the openwater. Noiselevels tend to enceestimation is a difficultproblem and is not definedtheoretically for beinsensitive to windspeed/sea state, showing a typicallevel of 83 dB re: 1 transientbroadband signals. Mean-square-coherence is computed using pPae/Hzat 100Hz. Usually,the buoy nearest the coast shows the highest frequencyband averaging and is computedseparately for the directand spectrallevel, suggesting that coastalprocesses are perhapsthe principal ice-reflectedpropagation paths. Coherence is found to fluctuatesignifi- sourceof low-frequencyambient noise in the Nordic Seas.Several mecha- cantlywith source-receiverrange, and acoustic frequency for bothacous- nismscould be responsiblefor the noise,one hypothesis being the move- tic propagationpaths. The highestvalues of coherencedo not always mentof glaciersfrom the Greenlandice cap, at an elevationof = 2000m, occur at the lowest frequencybands. Inside the acoustichalf-channel, down towards the sea. Each glacier containson the order of a billion above300 m, coherenceis lessfor the ice-reflectedarrival. Changesin tonnesof ice or morewhich grindsits way overrough, rocky terrain.This coherence due to under ice reflection are smallest for the 100- to 200-Hz movementgenerates seismic waves in theground which could couple into octaveband, indicating less coherence loss due to underice scattering, but the oceanthrough the seabed:An interface(Scholte) wavemay propa- thechange is sometimes as high as 50%. [ Work supportedby theOffice of gatealong the seabottom and radiateacoustic energy into the watercol- Naval Technology.] umn. S104 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S104 3:55 a l-km-longvertical array. Range and depth errors were less than 1% and I wavelength,respectively [ E. Livingstonand O. Diachok,J. Acoust.Soc. 007. Resultsof matched-field processing oflow-frequency Arctic under- Am. 86 (1989)]. At thelowest frquency, 11 Hz, therange estimates and icedata. E. Livingstonand O. Diachok(Naval Research Laboratory, Washington,DC 20375-5000) gainwere within 1% andI dB,respectively, but the error in thedepth estimatewas large, greater than three wavelengths. Replica fields were computedbased on a measuredbut range-independent sound-speed pro- Matched-fieldprocessing oflow-frequency ( 11 to 23 Hz) highsignal- fileand a constantbottom depth, whereas there was a majorbathymctric to-noiseratio cw signals in anArctic (under-ice) environment has pro- featurealong the propagation path. Possible causes for errorin depth ducedhigh signal gain (within I dBof theoretical)with good range and localizationdue to environmentalmismatch in the computedreplica depthlocalization for frequencies of 17 and 23 Hz, ata rangeof 260km on fields at 11 Hz will be discussed. THURSDAY AFTERNOON, 30 NOVEMBER 1989 DIRECTORS' ROW 26, 3:00 TO 5:30 P.M. Meeting of Accredited StandardsCommittee S3 on Bioacoustics to be heldjointly with the TechnicalAdvisory Group (TAG) Meetingfor ISOfrC 43 Acoustics,IEC/TC 29 Electroacous- tics, and ISO/TC 108/SC4 Human Exposureto Mechanical Vibration and Shock L. A. Wilber, Chairman S3 422 Skokie Boulevard, Wilmette, Illinois 60091 StandardsCommittee S3 on Bioacoustics. The current status of standards under preparation will be discussed. In additionto thosetopics of interest,including hearing conservation, noise, dosimeters, hearing aids, etc., considerationwill be given to newstandards that might be needed over the next few years. Open discussion of committeereports is encouraged. The internationalactivities in ISO/TC 43 Acoustics,IEC/TC 29 Electroacoustics,and ISO/TC 108/SC4 Human Exposureto MechanicalVibration and Shockwill bediscussed. The Chairsof theTAGs for ISO/TC 43 ( H. E. yonGierke ). IEC/TC 29 ( V. Nedzelnitsky),and ISO/TC 108/SC4( H. E. yonGierke) will report on current activitiesof theseTechnical Committeesand Subcommittees.ISO/TC 43 and IEC/TC29 will next meetin Copenhagenduring April 1990.TC 108/SC4will meetin Milan in March 1990. S105 J. Acoust.Soo. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S105 THURSDAY AFTERNOON, 30 NOVEMBER 1989 ST. LOUIS BALLROOM G, 3:00 TO 5:15 P.M. SessionPP. Physical AcousticsVII: Nonlinear Acoustics Wayne M. Wright, Chairman PhysicsDepartment, Kalamazoo College, Kalamazoo, Michigan 49007 ContributedPapers 3:00 considered.The Khokhlov-Zabolotskaya-Kuznetsovequation is usedto deriveasymptotic results for the sumand difference frequency sound that PP1. A continuous-wavereflection bridge ultrasonic spectrometer is generatedby thenoncollinear interaction of focusedsound beams radi- constructed from off-the-shelf components.Ronald K. Sundfors ated from displacedGaussian sources. As in the caseof unfocusedpri- (PhysicsDepartment, Washington University, St. Louis,MO 63130) mary beams,there are two maincontributions to the nonlinearlygenera- A versatilecontinuous-wave reflection bridge ultrasonic spectrometer ted soundfield: pumped sound and scattered sound. Attention is devoted has beendeveloped. This spectrometerfeatures simple tuning of the to the effectsof focusingon the relativeand absolute levels of the pumped bridgeelement (commercial hybrid junction) for frequenciesup to 100 and scatteredwaves. Strong focusing causes the pumpedand scattered MHz, a widevariation of appliedacoustic power, simple computer con- soundfields to broadenand overlap,which increasesthe difficultyof trol to lock frequencyto any point of the acousticresponse, diode and identifyingthe scatteredsound. Numerical results are presented.[Work phasecoherent detection, sensitivity equivalent to thebest nuclear acous- supportedby ONR (CMD and MFH), IR&D Programof ARL:UT, and tic resonancespectrometers, and construction from off-the-shelfcompo- VISTA/STATOIL (JNT and ST). ] nents.Use of the spectrometeras a cw spectrometer,sampled-cw spec- trometer,and NAR spectrometerwill be discussed. 3:45 3:15 PP4. Nonlinear elastic responseof shockedfused silica. Y. M. Gupta and M.P. Conner (Department of Physics, Washington State PP2. Effects of boundaryconditions on the propagationand interaction University,Pullman, WA 99164-2814) of finite amplitude sound beams. JacquelineNazc Tj•tta (Applied ResearchLaboratories, The University of Texas at Austin, Austin, TX Large amplitude,combined compression and shearwave measure- 78713-8029and Departmentof Mathematics,The Universityof Bergen, mentshave been obtained in fusedsilica (Dynasi11000) shocked to a peak 5007 Bergen, Norway), James A. TenCate (Applied Research compresslyestress of 6 GPa. The in-material,particle velocity profiles are Laboratories,The Universityof Texas at Austin, Austin, TX 78713- analyzedto obtain the longitudinalstress•tensity relation for uniaxial 8029), andSigve Tj•tta (AppliedResearch Laboratories, The University strain and shear wave velocities in the shocked state. The shear wave of Texas at Austin, Austin, TX 78713-8029 and Department of velocitydecreases with compressionto a densitycompression of 7.6%, Mathematics,The Universityof Bergen,5007 Bergen,Norway) and thenincreases with compression.Bulk modulus,Poisson's ratio, and meanstress in the shockedstate were determinedfrom the experimental Weak nonlincarityin the propagationand interactionof real sound beamsin a losslessfluid is considered.Special emphasis is givento the data.Comparison of the shockand hydrostaticcompression results show effectsproduced by variousboundary conditions at thesound sources and that the fusedsilica displays reversible, shear-enhanced compaction. Al- otherbounding surfaces. Asymptotic formulas and numericalresults are thoughtwo of the threethird-order elastic constants (C• and C•,,) can presentedfor the secondharmonic, and for the scatteredsum and differ- becalculated from the presentdata, the difficultyin describingthe fused encefrequency sound generated by two harmonicbeams that intersectat silicaresponse in termsof third-orderconstants is discussed.A micro- an arbitraryangle. The resultsarc derived from a generaltheory presented scopicmodel to explainshear enhanced compaction is suggested.[Work earlier [see Naze Tj•tta and Tj6tta, J. Acoust. Soc. Am. 83, 487-495 supportedby AFOSR, Grant No. 87-0081 under the DOD/URI pro- (1988) ], whichis valid for any sourceseparation and amplitudcdistribu- gram.] tion. In situationswhere the parabolicapproximation is not legitimate ( largeangles, broad beams), properly accounting for theboundary condi- tions may bc crucial. [Work supportedby the IR&D program of ARL:UT, ONR, and V1STA/STATOIL, Norway. ] 4:00 PPS. Nonlinear soundscatterlug of crossedfocused beams in the presence of turbulenceAngular measurements. Murray S. Korman (Department of Physics,United StatesNaval Academy, Annapolis, MD 3:30 21402) PP3. Effects of focusingon the scatteringof soundby sound.Corinne Experimentsare performedinvolving the interactionof mutuallyper- M. Darvennes, Mark F. Hamilton (Department of Mechanical pendicularcrossed, ultrasonic beams overlapping (at a commonfocal point) and interacting in the presenceof turbulencegenerated by a Engineering,The University of Texas at Austin, Austin, TX 78712- d = 0.64-cm-diamsubmerged water jet (Re = 8.5X 10s). Measurements 1063), JacquelineNaze Tj•tta, and SigveTjrtta (Applied Research Laboratories,The Universityof Texasat Austin, Austin, TX 78713-8029 of the radiatedsum-frequency intensity, Doppler shift, and-spectral andDepartment of Mathematics,The Universityof Bergen,5007 Bergen, broadeningare recordedfrom the scatteredspectra. In the absenceof turbulence,no sumfrequency can be detected. The two cw primarybeams Norway) • = 1.9 MHz, f2 = 2.1 MHz) are generatedby individual transducer The effectsof absorptionon the scatteringof soundby soundwere unitsof 15.2-cmfocal length. The receiver(a circularplane array located discussedby the authorsat the previousmeeting [J. Acoust.Soc. Am. outsidethe interactionregion) is positioned with its axisperpendicular to Suppl. I 85, S5 (1989) ]. In the presentpaper, the effectsof focusingare the jet at a distanceof 30d from the nozzle (the interactionregion). Pre- $106 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S106 viously,very good theoretical predictions of meanflow and rms turbulent ationpredictions. Finally, it is shownthfit there is an unexpectediso- velocitywere determined from the Dopplershift and spectral broadening morphismbetween the force fields in rectangularand cylindrical position- datafrom a similarcrossed-beam scattering experiment [M. S. Korman ing modes. andR. T. Beyer,J. Acoust.Soc. Am. 84, 339-349(1988); 85, 611-620 (1989)]. Theprimary beams were generated by circularplane array ele- ments.Results suggested that the interaction involved an averaging across a largeregion of flow.The localizedcrossed-beam interaction region al- lowsturbulent flow propertiesto be predictedwith reasonablespatial 4:45 resolution.[Work supportedby the NavalAcademy Research Council.] PP8. On supercriticalincidence of a finite-amplitudeplane dilatational wavein an elasticsolid. Kun-Tien Shuand Jerry H. Ginsberg(School of MechanicalEngineering, Georgia Institute of Technology,Atlanta, GA 30332) 4:15 An earlieranalysis [ K. T. Shuand J. H. Ginsberg,J. Acoust.Soc. Am. Suppl. I 85, S139 ( 1989} ] describednonlinear reflection and refraction PP6. Reflectionof a plane finite-amplitudepulse from a hard surface: phenomenaof a finite-amplitudedilatational wave at subcriticalincidence Amplificationfactor comparison between analytic and simulation. Victor ona planeinterface between two bonded solids. The present work extends W. Sparrow(Department of Electricaland ComputerEngineering, theearlier description to caseswhere the angleof incidenceexceeds the Universityof Illinois,Urbana-Champaign, IL 61801and USA-CERL, criticalvalue. The incomingwave is assumed to originatefrom the slower P.O. Box 4005, Champaign,IL 61824-4005)and RichardRaspet medium,so two critical angles exist, associated with evanescence of either (Departmentof Physicsand Astronomy, The University of Mississippi, the traosmiteddilatational or shearwave. The finite-amplitudeversion of University,MS 38677) Snel!'slaw indicates that the dependence of thephase speed of theincident The amplificationfactor for a soundwave impinging on a hardwall is waveon its instantaneousamplitude induces, in the caseof evanescent the ratioof the acousticpressure near the wall to the acousticpressure at waves,fluctuations in thephase velocity parallel to theinterface and in the thesame position with thewall absent. Linear acoustics gives the familiar decayrate normal to the boundary.This effectmirrors the fluctuationsin wall amplificationfactor of 2, pressuredoubling. However, the factoris thetransmission and reflection angles of propagativewaves. A numerical greaterthan 2 for finite-amplitudewaves. For regularreflection, as op- algorithmis developed to evaluatereflected or transmittedwaveforms at a posedto Muchreflection, the factor has been predicted in theideal gas specifiedfield point. In the specialcase of incidenceclose to a linearcriti- caseanalytically. In this presentationthe wall amplificationfactor pre- cal angle,the finiteamplitude Snell's law indicatesthat thecorresponding dictedby a nonlinearacoustics finite-difference simulation will be com- wavefluctuates between propagative and evanescent properties within a paredwith the analyticresult of Pfriem.A shortvideotape will demon- singleperiod. [Work supportedby NSF.] stratethe simulationvia animation.[Computations performed at the NationalCenter for Supercomputing Applications, Champaign, IL.] 5:00 4:30 PP9. The role of higher-order acoustic nonlinearities in NDE. Laszlo Adler, Peter B. Nagy, Paul McGowan, and David Rypien PP7. Force anisotropyin resonant acoustic levitators. M. Barmatz, (Departmentof WeldingEngineering, The Ohio StateUniversity, G. Archi, and JudithL. Robey (Jet PropulsionLaboratory, California Columbus, OH 43210) Instituteof Technology,Pasadena, CA 91109) The acoustoelasticmethod is a well-establishedway to obtain third- In general,the forcespositioning a samplein an acousticlevirotor order elastic constants in bulk materials. Modifications are needed to dependon direction,which is to saythey are anisotropic. The anisotropy extendthis technique to obtainhigher orders, e.g., fourth-order elastic of the acousticforce for severalclasses of levitotingmodes in rectangular constants,or to increaseaccuracy to adopt the techniquefor very thin andcylindrical resonators has been investigated. The degreeofanisotropy layers,e.g., adhesive bonds. These higher-order nonlinearities are separat- dependson the chamber geometry, mode that isexcited, and properties of ed from the strongerfirst-ordernonlinearities (associated with the third- the sample.The forceanisotropy has been studied as a functionof these order elasticconstants) by the appropriatechoice of polarizationand variablesfor singlemode levitators. The forcefield anisotropy is evaluated harmonicanalysis of dynamicacoustoelastic results. It hasbeen shown for smalldisplacements from equilibrium. Also, the anisotropicproper- that inclusionof the fourth-orderelastic constants will lead to strength- tiesfor the maximum positioning forces in theresonator that maintain the relatedmechanical properties of theadhesive bond. [This work was sup- samplewithin the acousticpotential well havebeen studied. Optimum portedby the Center for NDE operatedby the Ames Laboratory, USDOE chamberdimension ratios for maximizingthe latter forcefields will bc for the Air ForceAeronautical Laboratories/Materials Laboratory under presented.Experimental measurements verify the theoreticaloptimiz- Contract No. SC-89-168 with Iowa State University.] S107 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S107 THURSDAY AFTERNOON, 30 NOVEMBER 1989 ST. LOUIS BALLROOM BC, 4:30 TO 5:50 P.M. SessionQQ. PhysiologicalAcoustics V and PsychologicalAcoustics IX: Central Institute for the Deaf, S. Richard Silverman Lecture in Hearing and Deafness Richard G. Stoker, Chairman CentralInstitute for theDeaf, 818 SouthEuclid Avenue, St. Louis,Missouri 63110 Chairman's Introduction---4:30 Invited Paper 4:35 QQ1. Acousticinjury of the innerear: Morphological and physiological perspectives. M. CharlesLiberman (Departmentof Otolaryngology,Harvard Medical School and Massachusetts Eye and Ear Infirmary,243 Charles Street, Boston,MA 02114) Of the manyapproaches to the studyof acousticinjury, one of the mostfruitful has been to viewthe functionalchanges in the responseproperties of singleauditory-nerve fibers. Since each cooblear afferent contactsonly one inner hair cell, single-fiber responses provide a functionalwindow onto a restrictedportion of the cochlearduct. Through the useof intracellularlabeling, which allows the cochlearorigin of selected afferentsto beprecisely specified, correlations between structural and functional changes have been made on the single-celllevel. Correlations so obtained have revealed the functionallyimportant structural changes underlyingacoustic injury of theinner ear. Morphological data from both the light- and electron-microscopic levelssuggest that most sound-induced permanent threshold shifts can be accounted for quantitativelybased on the lossof sensorycells and/or the damageto theirstereocilia. Of all the structuresin the innerear, the stereocilia,and especially the "rootlets"that anchor them to thetops of thehair cells, appear to be themost vulnerableto permanentdamage from acoustic overstimulation. The structural changes underlying reversible thresholdshifts are clearlydifferent from thoseunderlying permanent loss, but exactlywhich structural changesunderly the temporary losses are much less clear. Structure-function correlations in damagedears havealso helped define the contributions ofinner versus outer hair cells to theresponse properties of auditory- nervefibers. Damage to innerversus outer hair cellsresults in strikinglydifferent abnormalities in neural tuning,spontaneous activity, and maximal discharge rates. The nature of thesedifferences suggests that audi- toryperceptual abilities would be very different given threshold shifts of identical severity but differing balance of damageto the two classesof sensorycells. $108 d.Acoust. $oc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S108 FRIDAY MORNING, 1 DECEMBER 1989 DIRECTORS'ROW 43, 8:00TO 9:45A.M. SessionRR. PhysicalAcoustics VIII: Ultrasonics,Photoacoustics, Optoacoustics, and Thermoacoustics Anthony A. Atchley, Chairman PhysicsDepartment, Naval Postgraduate School, Monterey, California 93943 ContributedPapers 8:00 tionac/f 2 vsf •½,6yields a straightline as predicted by thedynamic RR1. The use of reciprocity to calibrate a resonantphotoacoustic scalingtheory of Ferrelland Bhattacharjee. The criticalamplitudes of the measurement.J. D. Maynard and Chang Yu (The PennsylvaniaState thermalexpansion and specific heat have been calculated using the two- University, University Park, PA 16802) scalefactor universality relation. The adiabaticcoupling constant g is calculatedand comparedto the experimentalvalue. The experimental While the photoacousticeffect provides a powerfulmeans of measur- valuesof a/ac for nitrobenzene-n-hexaneare comparedto the scaling ing verylow opticalabsorptions with pulselasers, obtaining an accurate functionF(eo*) andshow a goodagreement with thetheory. The velocity calibrationhas been a problem.However, in a resonantphotoacoustic for thesystem at thecritical concentration above the critical temperature measurement,a reciprocity calibration (which does not require replacing appearsto linearlydecrease with increasingtemperature. transducers)is readilyaccomplished. All that is requiredis that a reversi- ble transducerbe present in additionto the photoacoustic(laser) excita- tionand the usual photoacoustic receiver. By makingonly electrical mea- surementsusing alternately the laserand the reversibletransducer for excitation,the absoluteacoustic levels in the samplemay be determined fromthe transducer voltage. By alsomeasuring the quality factor (Q) of 8:45 the resonancefrom the frequencyresponse, the powerdeposited by the laseris easilydetermined. A separatemeasurement of the incidentlaser RR4. Observation of optoacousticamplitude in CS2 at high input poweryields the opticalabsorption coefficient. [Work supportedby the energies. Stanley A. Cheyne (School of Mechanical Engineering, Office of Naval Research.] GeorgiaInstituteofTechnology,Atlanta, GA30332) andHenryE. Bass (PhysicalAcoustics Research Group, The University of Mississippi, University,MS 38677) The optoacousticamplitude as a functionof laserpulse energy has 8:15 beenmeasured. A nitrogenlaser emitting UV (337 nm) pulses800 ps in durationwas uscd to exciteliquid carbondisulfidc (CS2), whichstrongly RR2. A newresonant photoacoustic technique for measuringvery low absorbsthe UV radiation.Measurements indicate an opticalabsorption opticalabsorption in glassesand crystals. Chang Yu andJ. D. Maynard coefficientof 370cm • implyingan optical penetration depth of 27/•m. (The PennsylvaniaState University, University Park, PA 16802) The spatialprofile of thepulse was an ellipseat thefocal point with dimen- sionsof 330)<90 •am. The energycoupled with the short penetration Oneof themost sensitive methods for measuring low optical absorp- depthresulted in a highenergy density (532 J/cm3). This was varied by tion in glassesinvolves the photoacousticeffect. Conventional methods placingglass microslides in thebeampath. The optoacousticamplitude as usehigh-power pulse lasers and attached piezoelectric transducers to ob- a functionof energyexhibited changes in slopeas the liquid in the focal tain detectableacoustic signals, but thismethod is limitedby noiseand volumeundcrwcnt vaporization. The experimentalobservations were ex- scatteredlight at the transducer.A newtechnique has been developed plainedin termsof an energy-dcpendentcoefficient of thermalexpansion usinga cw laser,which, although it hasless power, may be modulated at and optical absorption.[Work supportedby the Officeof Naval Re- the acousticresonant frequency of the sampleand thusgenerates an search.] acousticsignal amplified by thequality factor (Q) of theresonance. Since Q'sof 105to 106are readily obtained, a cw laser with only 0.01% of the energyof a pulselaser may be used to obtainseveral orders of magnitude improvedsensitivity. To increasethe Q andthus increase the sensitivity, themeasurements aremade in vacuum,the sample supports (three sa- phirespheres) are located at nodallines with computer-controlled posi- tioners,and the transducersare noncontact interdigital capacitors that 9:00 sensethe motionof a nearbydielectric sample or the fringefield of a RRS. The measurement of thermoacoustic phenomena using nearbypiezoelectric sample. The noncontacttransducers also eliminate thermoacousticcouples. Anthony A. Atchlcy, Thomas J. Hofier, and the problemof scatteredlight. [Work supportedby theOffice of Naval Chia-ning Ao (Physics Department, Naval PostgraduateSchool, Research.] Monterey, CA 93943) The resultsof measurementsof thermoacousticallygenerated tem- peraturegradients in thermoacousticcouples (TACs) subjectedto acous- tic standingwaves are reported. The valueof the temperaturegradient is a 8:30 functionof the acousticpressure amplitude, mean gas pressure, type of RR3, Critical behaviorof the ultrasonicattenuation and velocityand gas,TAC construction,and its positionin the standingwave. Measure- shear viscosity for the binary mixture of nitrobenzene-n-hexane.I. mentswere made with a computer-controlledapparatus for drive ratios R. Abdelra•iq, S.S. Yun, and F. B. Stumpf (Department of Physics (the ratio of acousticpressure amplitude to mean gas pressure)from and Astronomy,Ohio University,Athens, OH 45701-2979) approximately0. 1% to 2.0%, in argonand helium having mean pressures fromapproximately 0.1 to 0.3 MPa, for threedifferent TACs asa function Ultrasonicvelocity and absorption and shear viscosity as a functionof of their positionsin the standingwave. The resultsare comparedwith temperatureand frequencyare reportedfor the binarymixture nitroben- predictionsbased on a theoryby Wheatleyet al. [J. Acoust.Soc. Am. 74, zene-n-hexanein the homogeneous phase above T c . The observedabsorp- 153-170(1983) ]. Threedistinct regions of behavior are apparent over the S109 d. Acoust.Sec. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America S109 rangeof drive ratios investigated. For drive ratios less than approximately portedby theDepartment of Interior'sBureau of Minesunder Contract 0.4%, thereis overallgood agreement between theory and measurement. No. J0134035through Department of Energycontract No. DE-ACO7- For drive ratiosbetween approximately 0.4% and 1%, the agreement ?6IDO 1570.] diminishesalmost linearly with increasingdrive ratio. For drive ratios greaterthan approximately 1%, irregularitiesappear in the temperature differencedata seriesand the discrepancybetween theory and measure- mentgenerally remains approximately constant, although some varia- 9:30 tionsoccur. [ Worksupported by ONR andthe NPS ResearchProgram. ] RR7. Ultrasonic behavior in an amorphous superionic conductor {Agl}x{AgPO3}•_x. Jiankai Hu, Senkui Zhang, and Qianlin Zhang (Departmentof Electronics,University of Scienceand Technologyof 9:15 China, Hefei, Anhui, People'sRepublic of China) RR6. Laser ultrasoMe measurements in highly porous materlals: Amorphoussuperionic conductor (AgI) x(AgPO3)• x seriessam- Sinterlngof ceramics.K. L. Telschow,J. B. Walter, and G. V. Garcia pleshave been obtained by meansof the liquid-nitrogenrapid-quenching technique.In the seriesof glasses,for Agl with moleconcentration of (Idaho National EngineeringLaboratory, EG&G Idaho, Inc., Idaho x = 0.45, 0.5, 0.55, and 0.6, the attenuationand velocityof longitudinal Falls, ID 83415-2209) waveswere measured by an ultrasonicpulse-echo method, in thetempera- The useof lasersfor bothgeneration and detectionof elasticwaves in ture range100-300 K at frequenciesof 5 and 10.5MHz. The temperature materialshas opened the possibilityof real-timemonitoring of material of the attenuationpeak fell with increasingAgI contentand decreasing elasticparameters during processing. At the INEL, a laserDoppler inter- frequency.In addition,the amplitudesof theattenuation peaks increased ferometer has been constructed that allows direct measurement of elastic significantlywith increasingAgl content,despite the fact that the total wavesfrom rough and/or optically diffuse surfaces. This apparatus, cou- concentrationofAg + ionsis about constant for thisseries of samples.The pledwith a pulsedlaser for elastic wave generation, is currently being used soundvelocity of the samplex = 0.5, at frequency5 MHz, decreased for totallynoncontacting, real-time, monitoring of thesintering of ceram- monotonouslywith increasingtemperature and had a small anomalyat ics.Results will bepresented on the densification of materialsas a function 204 K. The soundvelocity of the glassesdecreased slowly as x wasincreas- of sinteringtime and temperature.Both the longitudinaland shearwave ing at roomtemperature. The experimentalresults were fitted well to the velocitiesin thesematerials have been found to vary approximatelylinear- unifiedtheory of low-frequencyfluctuation, dissipation, and relaxation ly with the densificationup to nearly total theoreticaldensity. Mecha- process,i.e., the infrareddivergence response theory proposed by Ngai nismsof the generationand detectionof theseelastic waves, as well as [K. L. Ngai,Comments Solid Phys. 9, 127 ( 1979);also, Phys. Rev. B 20, propagationin porousmaterials, will be discussed.[This work wassup- 2475 (1979) ]. [Work supportedby NSF of China.] FRIDAY MORNING, 1 DECEMBER 1989 ST. LOUIS BALLROOM B, 8:30 TO 11:50 A.M. SessionSS. PsychologicalAcoustics X: Central Processesin Auditory Detection,Discrimination, and Recognition BiasEspinoza-Varas, Chairman Universityof OklahomaHealth SciencesCenter, Oklahoma City, Oklahoma73190 Chairman's Introduetion-•8:30 Invited Papers 8:40 SS1. Baslesensitivity on vowel and consonant continua: A functionof phonemecategories. Bert Schoutenand Arian van Hessen(Research Institute for Languageand Speech,University of Utrecht,Trans 10, 3512JK Utrecht, The Netherlands) The relationbetween labeling and discriminationof stimuli drawn from a short-voweland a plosive- consonantcontinuum was investigated. Stimuli were generated by interpolationbetween the spectral envelopes (derivedwith cepstralaleconvolution) of naturallyspoken syllables. The taskswere phoneme labeling, fixed androving same•lifferent discrimination with or withoutfeedback, and fixed and roving 2IFC discrimination with feedback.The resultsindicate that, with natural speechsounds, labeling precedes the discrimination response,even in fixeddiscrimination, which is supposedto measurebasic sensitivity. Equivalence between labelingand discrimination was strongest with consonants,which agrees with the old categoricalperception literature.Varying ISI had a smalleffect on fixedsame-different discrimination; at 100ms, performancewas relativelyhigh, and significant category peaks had disappeared; this could suggest basic sensitivity, but labeling performancewas even higher. The effect of ISI on 2IFC discriminationwas the reverse: a longISI of 2000ms improvedperformance considerably, especially at thecategory peak. It seems,therefore, that with speech2IFC is a labeling-orientedtask: It improveswith time,particularly in a fixed-discriminationparadigm. Sl 10 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America Sl 10 9:05 $$2. A performance-orientedapproach to thehearing of complexsounds: Nonspeech and speech. Charles S. Watson(Department of Speechand Hearing Sciences, Indiana University, Bloomington, IN 47405) Twogeneral questions have been addressed with experiments using word-length tonal sequences, or "pat- terns":to whatdegree can the discriminabilityof complexauditory patterns be predictedfrom listeners' performancewith simplerstimuli, especially single tones presented in isolation?and, (2) How do listeners' abilitiesto extract information from complex nonspeech sounds compare to theirabilities to resolvethe details ofthe waveforms ofspeech? The answers appear to be consistent with the results of comparable investigations conductedfor thepast half-century in thefield of vision.Very little degradationin sensitivityor resolving powerfor componentsofcomplex sounds can be attributed to thecomplex stimulus context per se, aside, of course,from temporal masking. When patterns vary from trial to trial, selectiveattention and accurate dis- criminationperformance can be achieved only when prior experience has made them so familiar that process- inghas become "automatic." Listeners' remarkable resolving power for speech sounds can readily be imitated by overtrainingthem on thediscrimination of at leastsome nonspeech sounds. 9:30 SS3.Auditory profile analysis. David M. Green(Psychoacoustic Laboratory, Department of Psychology, Universityof Florida, Gainesville,FL 32611) Dataare reported on theability of listenersto discriminatechanges in thepower spectra of complex acousticsignals. Since randomizing both the phase spectrum and the overall level of the signal on each presen- tationdoes not change detection performance, it is knownthat detection is notbased on thesignal's phase spectrumor theabsolute level of energy at a particularfrequency location. The discrimination must be based on a changein theshape of thepower spectrum of thesignal, what is calledprofile analysis. The most critical featureof profile analysis isthat detection of a changeat one spectral location depends on energy at veryremote spectrallocations. For example,the detection of an incrementin theintensity of a componentat 1000Hz is influencedby thepresence or absenceof energymore than 2 oct away.Recent experiments show that these effectsmay depend on whetherthe standard power spectra are fixed or changingover successive trials, since spectrawith randomlychanging levels do not showsuch effects. Finally, some recent experiments will be discussedthat provide estimates of thespectral weights employed by profile listeners, using a techniquedevel- opedby Dr. BruceBerg. {This research was supported by NIH andAFOSR. ] 9:55 SS4.Relation between comodulation masking release and auditory grouping. Joseph W. Hall, Ill andJohn H. Grose(Department of Surgery, Division of Otolaryngology,University of NorthCarolina Medical School, ChapelHill, NC 27599) CMR wasinvestigated in conditionswhere some flanking hands were comodulated with theon-signal band,but other flankingbands ("codeviant" bands) were comodulatedbetween themselves, but not with the on-signalband. Past research has indicated that CMR isoften substantially reduced when codeviant bands are present.It ispossible that the disruptive effects oleodeviant bands can be reduced by factors related to auditory grouping.In onecondition, 20-Hz-wide comodulated bands were present at 400, 600, 800, 1000,1200, 1400, and 1600Hz ( 1000-Hzsignal frequency). When two codeviantbands were added at 900 and 11{30Hz, CMR wasreduced from 16 to approximately 4 dB. Progressively more codeviant bands were then added, interspersed with thesix comodulated bands. The rationalewas that as more codeviant bands were added, they were more likely to be groupedseparately from the on-signalband, and their disruptiveeffects would, therefore, be reduced.A secondmanipulation was to gatethe codeviantbands differently from the comodulatedbands: If thecodeviant bands are givena uniquegating pattern, they should be lesslikely to be groupedwith the comodulatedbands. Both of theabove manipulations significantly reduced the disruptive effects of codeviant bands.The resultssuggest that onestage of CMR analysisapparently takes place subsequent to a stageof auditorygrouping. [Work supportedby the U.S. Air Force,AFOSR. ] 10:20-10:35 Break S111 J.Acoust, Soc. Am. Suppl. 1,Vol. 86, Fall 1909 118thMeeting: Acoustical Society ofAmerica S111 10:35 SS5. Factorsin detectionunder uncertainty. Ervin R. Hafter and Robert S. Schlauch(Department of Psychology,University of California,Berkeley, CA 94720) An especiallyeffective tool for examiningcognitive factors in hearingis the studyof maskingunder conditionsof uncertainty.Past studiesof frequencyuncertainty and its alleviationwith cuesleads to the proposalthat the role of attentionin detectionis to governthe trade-offbetween the numberof monitored bandsand their effectivebandwidths. In an extensionof that work, attemptsto measurethe bandwidths directlyand to understandthe nature of cuingare made, whether it isa bottom-upprocess, as in theringing of sensoryfilters, or top-down,as in theselection of theappropriate filter. Two distinct approaches are offered. In oneapproach, a modified"probe" technique presents a cueof either 1, 2, or 4 toneswhich includes, on most trials,the frequency of thesignal; on probetrials, the signal is some fixed percentage (other than 100% ) of the frequencyof onethe cues. Performance based on datacollapsed across equal percentages support the conten- tionthat changes occur in botheffective number and bandwidth. Additional studies examine cues that require higher-orderprocessing such as computation of musicalinterval or fundamentalpitch. The effectiveness of thesecues supports the notion of top-downprocessing. [Research supported by Grantsfrom NINCDS.] 11:00 SSr. Centralprocesses in the perceptionof pitchin simultaneouscomplex tones. John G. Beerends(P.T.T. Research,Ncher Laboratories,P.O. Box 421, 2260 AK Leidschendam,The Netherlands) Pitchidentification of twosimultaneous complex tones is moredifficult than pitch identification of a single complextone. In thistalk, the morecentral processes that are involvedin the perceptionof simultaneous complextones will be discussed.Experiments are reportedin whichsubjects had to identifytwo missing fundamentalfrequencies. A first set of experimentsuses simultaneous two-tone complexes with different distributionsof thefour partial over the ears. It isshown that subjects can identify the two missing fundamental frequenciesand that the identificationresults only weaklydepend on the mannerof distributingthe four partialsover the ears. For somesubjects, the identification results showed that the more salient pitch can inhibit to someextent the identificationof the lesssalient one. This "centralinterference" is mostprominent when the partialsof the two complextones are widely separated. In a secondset of experiments,the influenceof signal durationon the virtualpitch of singleand simultaneouscomplex tones is investigated.The resultsfor single complextones allow the formulationof a stochasticpitch model in whicha centralprocessor derives noisy subharmonicsfrom noisyrepresentation of thepartials. For simultaneouscomplex tones, the behavioris more difficultto modelbecause of the morecentral processes that are involved. 11:25 SS7. Mechanismsof abstractionin auditory processing.Ernst Terhardt (Institute of Electroacoustics, TechnicalUniversity, D-8000 Munich 2, FederalRepublic of Germany) The naturalpurpose of anysensory system is acquisition of informationon external processes and objects, i.e., abstractionof the stimulus. Abstraction, in turn, is typically dependenton decisionsthat are basedon internalknowledge. To getmore insight into auditory information acquisition--including perception of speech andmusic--it is thus apparent that particular attention should be paid to decisionprocesses on any level of the auditorysystem. Inspecting the auditorysystem from that pointof view revealsthat both its physicaland neurophysiologicaldesign and its psychophysical behavior indeed exhibit many cues that are typical of abstrac- tion mechanisms.For instance,extraction of individual "pure tones" from a continuoussound spectrum requiresa decisionmechanism; therefore, the formationof spectralpitches in the auditoryperiphery may be regardedas a firststep of abstraction. On thebasis of a numberof experimental data and theoretical arguments, the followingquestions will be discussed:( 1) How is the hierarchyof abstractionorganized? (2) Where is the "sensoryknowledge" stored? ( 3 ) What is theorigin of thatknowledge? (4) Whichstrategies of experimental researchappear to be appropriate? S112 J.Acoust. Soc. Am. Suppl. 1, VoL 86, Fall 1989 118thMeeting: Acoustical Society of America $112 FRIDAY MORNING, 1 DECEMBER 1989 ST. LOUIS BALLROOM A, 8:30 A.M. TO 12:09 P.M. SessionTT. SpeechCommunication IX: Articulation: Measurementsand Models Ralph H. Ohdc, Chairman VanderbiltUniversity, School ofMedicine, Division ofHearing and Speech Science, Box 562, Station 17, Nashville, Tennessee 3 7232 ContributedPapers 8:30 activityin the tongue,and to generatearea functions of the vocaltract in order to specifydirectly the mappingfrom muscleactivity to acoustic TY1. Recovery of tongue function for speechand swallowingin partial output,As a firststep in a finiteelement description, a referencemodel of glossectomy patients. Sandra L. Hamlet, Robin Patterson, and the tongueis constructedthat can be adjustedto differentindividual Susan Fleming (Department of Otolaryngology, Wayne State shapesand structures of thetongue muscles, based on anatomical studies. University,Detroit, MI 48201) A schematictopology of thetongue is generated by separatinga region of Most previousresearch on glossectomyspeech has been cross section- a Cartesianspace into simple geometric shapes representing different tis- al--patients weretested at only one point in time, usuallyseveral months suecomponents. A mappingfrom this descriptivespace to a curvilinear or yearspost surgery. This studyof oral cancerpatients was designed to coordinatesystem using B-splines in three variablesis specifiedto gener- provideinsight into the developmentof compensatoryarticulation and atea realisticinitial shape of thetongue. The orientationof musclefibers swallowingpatterns in the early postsurgery and subsequentradiation can be determinedby normalizedgradient fields in the descriptivespace. therapy period. Synchronizedvideofluoroscopic and acousticalrecord- ingsof speechand swallowingwere madeat specifiedtimes during and after cancertreatment. Tongue mobility relative to thejaw was greater postradiation treatment compared to postsurgery. Expected co-articula- 9:06 tory variation in secondformant frequencyof vowels,as a function of consonantcontext, did not reappearuntil 6 monthspost treatment. TY4. Lip protrusion in sibilant production. Alice Faber (Haskins Tongueactivity for theoral phaseof swallowingshowed similar recovery Laboratories, 270 Crown Street, New Haven, CT 06511 ) of differentialpatterns related to bolusvolume. Data on rule-governed subtletiesof tongueactivity related to detailsof the speechor swallowing Whileanecdotal reports of lip protrusioncontributing to sibilantpro- contextprovide a sensitivemeasure of reestablishmentof naturalnessin ductionin a varietyof languagesexist, little attentionhas been paid to tonguefunction. quantifyingits contribution.In the presentstudy, lip positionwas moni- tored usinga modifiedSELSPOT optoelectronictracking system, and locationof oral constrictionwas simultaneously monitored by meansof a RION artificialpalate. Subjects (speakers of Catalan, English,German, and Italian) produced50 tokenseach of 6 (8) symmetricalVCV utter- 8:42 ances,where V rangedover/iau(ii)/and C over/s•/.As expected,/s/ has a more anterior constriction than/•/, with the constriction for/usu/ Tr2. Cross-sectionaltongue shapesand palatal contoursduring sibilant beingthe mostanterior; and, thereis morelip protrusionfor/•/than for and lateral consonants.Maureen Stone(Department of Rehabilitation /s/, exceptfor in the/u u/context. Acrossutterance types, the combina- Medicine, Building 10, Room 6S235, N.I.H., Bethesda,MD 20892), tion of lip protrusionand constriction location is a betterpredictor of the LawrenceJ. Raphael,and Alice Faber (Haskins Laboratories,270 frequencyof a sibilant'scentroid than is eitherparameter alone; within Crown Street, New Haven, CT 06511 ) utterancetypes, lip protrusionand aperture are, in general,better predic- This presentationwill examinethe relationshipbetween tongue shape torsof centroidfrequency than is constriction location. [Work supported and palatalcontour during the productionof/s/,/.f/, and/1/. A combi- by NIH Grant NS-07237. ] nation of elcctropalatoraphy(EPG) and ultrasound(US) was usedto comparetongue-palate contact patterns with theirrelated cross-sectional tongueshapes. The phonemes/s/, /.f/, and/1/arc interestingto study becausethey are physiologicallydemanding. They necessitatedistinct 9:18 posteriortongue shapes to channelthe air anteriorly (sibilants)or lateral- ly (/l/). In addition,these phoneroes exhibit specific palatal contact pat- •FS. The front-cavity/F2' hypothesis tested by data on tongue terns that result in their uniqueacoustic waveforms. Another effectof movements.David J. Broad (1627 Bath Street, Apartment 3, Santa tongue-palatecontact, which is usuallyoverlooked, is the resultantbrac- Barbara, CA 93101) and Hynek Hermansky (U.S. West Advanced ing of the tongue.With the anterior tonguebraced against the alveolar Technologies,6200 Quebec Street, Englewood, CO 80111) ridge,the posterior tongue is able to deforminto a varietyof cross-section- al shapes(concave for/s/, convexfor/1/). The hypothesisthat the secondspectral peak F2' resultingfrom per- ceptuallinear prediction ( PLP ) correspondsto the resonance of thefront cavityof the vocaltract is testedagainst data on tonguemovements for non-low-backvowels. The front cavity is representedfor this caseas a uniformtube of lengthL/. Scalingbetween movements of thetongue Ax 8:54 (rangeof a few centimeters)and of the vocal-tractconstriction Tr3. Construction of a three-dimensional tongue model. (range of about 10 cm) is representedby an articulatoryamplification Reiner Wilhelms (Speechand Hearing ScienceDivision, The Ohio State factora: ALl = aAx. If x isa tongue-surfacecoordinate parallel to the tract axis, then I/F2' = k 4- 4axle, where k is a constant and c is the University,Columbus, OH 43210) speedof sound.When x is the x coordinateof a tongue-dorsumpellet A three-dimensionalmodel for the investigationof the dynamic be- trackedby x-ray microbeam,the correlationsbetween l/F2' and x are haviorof the tongueis under study.Its purposeis to relate x-ray micro- 0.74 and 0.75 for two speakers'triple repetitionsof/wiw•juojil/. The beam-measuredmovements of pelletson the tonguesurface to muscle respectiveestimates for a are 5.5 and 5.6.The correlationsrise to 0.91 and S113 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S113 0.92 if x is a weightedsum of the x and y coordinatesof three tongue fixedframes "Say a Cawa"and "Say a waCa"with C = {fo/, /d/, /g/). pellets.Two predictionsare thus confirmed: ( 1 ) a linearrelation between Eachutterance was repeatexl three times in eachof threespeaking styles I/F2' and tonguemovements, and (2) reasonablevalues for a. A more rangingfrom an exaggeratedstyle intended to be "clear" (i.e., easierfor complexmodel of the front cavitymay be neededto handlelow back the hardof heatingto understand)to a relaxedstyle intended to be "con- vowels.[The authorsthank George Pap•un ofLos Alamos National Lab- versational."The audiorecordings were presented to hearingimpaired oratoryfor accessto his x-ray data.] listenersin a three-alternativeforced-choice stop consonant identification task to determinethe relativeintelligibility of the stopconsonants. By hypothesis,the clearestconsonants would be found in utterancesthat exhibitedthe greatestamounts ofarticulatory and acoustic change. How- ever,this patternwas not found.For utteranceswith syllable-initialcon- 9:30 sonants(e.g., "Say a gawa"), actualintelligibility as estimated by percen- TY6. Interarticulatoryrelationships in vowelproduction. Mona Lindau tagecorrect stop constant identification was inversely related to boththe andPeter Ladefoged(Department of Linguistics,UCLA, LosAngeles, talkers intent and measures of the extent of articulator motion. Correla- CA 9O024-1543) tional analysesusing both articulatoryand acousticmeasures to predict obtainedintelligibility scores will be presented.[Work supportedby In earlierstudies it hasbeen shown that, within a language(English), NIH. I thejaw andthe tongue interact in differentways for differentspeakers to producedifferences between vowels. These studies were, however, limited to front vowels.Using the x-ray microbeamsystem, articulatory move- mentswere recorded from pellets placed on the lips,mandible, and the tongue,in fivespeakers saying utterances illustrating all theEnglish vow- 9:54 els.The resultsshow some rather complex relationships. For example,for somespeakers tongue height, as measuredby the heightof the tongue TTS. Phase transitions in speech production and their perceptual dorsum,does not alwaysdistinguish between the highfront/i/and mid consequences.Betty Tuller andJ. A. Scott Kelso(Program in Complex front/e/(e.g., in deeand day); but these vowels are distinguished by the Systemsand Brain Sciences,Florida Atlantic University,Boca Raton, FL heightof themandible. The height of thedorsum does separate the tense 33431) /i/and thelax/t/(e.g., in deeand did) andthe back/u/and/o/(e.g., in Previouswork showsthat the relativephasing of articulatoryevents doand doe). The resultsalso show the degreeof consistencywithin indi- varieslittle acrosstwo levelsof speakingrate and stress. But, one may ask, vidualsin favoringcertain trading relationships between the lips, the man- varieslittle with respectto what?A methodologyisrequired that differen- dible,and the tongue. tiatesarticulatory patteros as they evolve in time.With thisaim, subjects wereinstructed to say/ip/or/pi/repetitively, at increasingspeaking rates,while monitoringtheir glottaland lip movements.For/pi/, the observedinterarticulatory phase relationships did notdepend on speaking rate. For/ip/, relativephasing often changed markedly, suggesting that lossof stabilityunderlies articulatory change (a phasetransition). In two follow-upperceptual experiments, listeners appeared to judgethe cons• TT7. Articulatory and acoustic analys'mof consonantsdiffering in nant'ssyllable affiliation on the basisof interarticulatorrelative phase, intelligibfiity. H. Timothy Bunnell (A. I. dupont Institute, 1600 and did soin a "categorical"manner. These findings, interpreted in light RocklandRoad, Wilmington,DE 19899) of a theoreticalmodel based on synergeties[ Kelso et al., PhysicsScripta Articulatoryand acoustic recordings of a singletalker producing mul- 35, 79-87 ( 1987) l, mayhelp resolve controversies concerning whether or tiple repetitionsof nonsenseutterances were made at the Universityof nota particularvariable is an articulatory"invariant." [Work supported WisconsinX-Ray Microbeam facility. The utterancesconsisted of the by NIH and NIMH.] 10:06-10:21 Break 10:21 obtained.Results demonstrate that coordinativetiming relations among the multiplearticulators are systematically scaled as a functionof move- TT9. Speech movement coordination: Oral-laryngeal interactions. mentspeed and phonetic context. Thus, consistent relative timing among Vincent L. Gracco (Haskins Laboratories, 270 Crown Street, New functionallyrelated articulators is a generalproperty ofspeeeh movement Haven,cr 06511} andAoders Lofqvist(Department of Logopedicsand coordination.It is hypothesizedthat such a constraintwhile simplifying Phoniatrics,University Hospital, S-221 85 Lurid,Sweden, and Haskins the motorcontrol process can providesufficient kinematic variability Laboratories, 270 Crown Street, New Haven, CT 06511 ) whilemaintaining the underlyingsynergistic organization. [Work sup- Kinematicinvestigations of speechmotor actionshave revealed ele- portedby NIH.] gantexamples of coordinativemotor control principles that reducethe degreesof movementfreedom. For example,recent studies of upperlip, lowerlip, andjaw movementsfor oral closurehave demonstrated a tem- 10:33 poralorganization similar to rhythmicmotor behavior such as locomo- tion.Consistent temporal coupling among multiple articulators is main- TT10. The kinematic of final lengthening in French. Janet Fletcher tained and systematicallyscaled following unexpectedmechanical (LinguisticsDepartment, 204 Cunz Hall, 1841Millikin Road,Ohio State perturbationsof thelower lip or experimentalmanipulations of thepho- University, Columbus,OH 43210-1229) and Eric Vatikiotis-Bateson neticenvironment. In order to generalizethese apparent eoordinative (Haskins Laboratories, 270 Crown Street, New Haven, CT 06511 ) principlesto includeother phonetic environments and articulatory sub- systems,the presentstudy investigated the adjustmentsunderlying lip, In a crosslinguistic study of lip-jaw motion, Vatikiotis-Bateson jaw, andvocal and vocal fold movements associated with various voiceless (1988) showedthat the kinematicpatterns associated with the nonfinal sounds.Movements of the upperlip, lowerlip, jaw, and vocalfolds were accented-unaccented contrast in French are similar to those found for the recorded.Kinematic measures,including relative timing amongjaw, up- stressed-unstressedcontrast in English.Opening movements generally perAowerlip, andglottal opening/closing (alevoicing) movements were havelonger durations, greater articulator amplitude, and higherpeak Sl 14 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society ofAmerica S114 velocitywhen accented than whenunaccented, In the currentstudy, the 11:09 sameFrench corpus is used to investigate final lengthening. Like English, TT13. In search of lingual stability. Eric Vatikiotis4Bateson(Haskins Frenchsyllable duration is longer at theedge of intonationalphrases than Laboratories, 270 Crown Street, New Haven, CT 06511) and in precedingsyllables. Yet, it is not clearwhat the articulatory correlates Maureen Stone(Department of RehabilitationMedicine, Room 6S235, of thisfinal lengthening are, nor to whatextent they are language-specific. Building10, NationalInstitutes of Health, Bethesda,MD 20892) A recentstudy of finallengthening in English(Edwards, Beckman, and Fletcher,in press)showed that the extra lengtheningof pitch accented In this study,commensurate data were obtainedfor tonguemotion syllablesthat are also phrase-final isaccompanied by lowerpeak velocities (coronaland sagittalultrasound scans), lingual-palatal contact patterns in thelatter part of thesyllable with no changein articulatoramplitude. (dynamicpalatography), and jaw motion(Selspot). Two subjectsrepeat- However,initial resultsfor Frenchindicate that lengtheningin phrase- ed/VCVCa/combinations of/s,.[,1,i,a,o/.To date, the effectsof conso- final syllables,which are alwaysaccented, is marked by higher artieulator nant identity on jaw motion and palatal contact patterninghave been amplitudeand peak velocity.Thus the kinematicsof final lengtheningis examined,along with the anticipatoryand carryovereffects of the three qualitativelysimilar to that of nonfinalaccented syllables. The conse- vowels.Initial findingson bothjaw heightand palatalcontact suggest that quencesfor an articulatoryand linguistictiming model of Frenchwill be the tongue-jawcoupling observed for sibilantproduction was modulated discussed.[Work supportedby Ohio StateUniversity, NSF, and NIH. ] systematicallyaccording to adjacentvowel identity. In contrastto/s/and /•/, tongue-jawcoupling for/1/appears to be minimalas shown by the large token-to-tokenspatiotemporal variability of jaw motion.Instead, there is a very stable,yet highly asymmetricalpattern of lingualcontact alongthe outer edge of theartificial palate. Analogous to thejaw's rolein sibilantproduction, this may serve as a stabilizingbrace for the tongueto 10:45 achievepostures specific to/l/production, whichshould be identifiable TYII. Remote and local effects of stresswithin and amongarticulatory from analysisoœ the ultrasounddata. These results are consistentwith a subsystems.Carol A. Fowler (Haskins Laboratories,270 Crown Street, task dynamic approach to interarticulatorcoordination [Saltzman New Haven, CT 06511 and Dartmouth College,Hanover, NH 03755), (1986) ]. [Work supportedby NIH.] Vincent Gracco, and Eric Vatikiotis-Bateson (Haskins Laboratories, 270 Crown Street, New Haven, CT 06511 ) Previousresearch on stresshas focused almost exclusively on articula- tory correlateslocal to a singletarget syllable within a singlearticulatory subsystem(e.g., the lip-jaw or the respiratorysystem). However, the 11:21 artieulatorycorrelates of stressare not local in either sense.Although a few studiesreport remoteeffects of stress(effects not confinedto the TT14. Palatography:A comparisonbetween custom-made and "flexible" stressedsegment) on the speechacoustics, articulatory studies have re- artificial palates for speechproduction measures.Nancy S. MeGart, portedthe effects of stressonly for singlesubsystems (e.g., oral, laryngeal, Koichi Tsunoda (Haskins Laboratories, 270 Crown Street, New Haven, and respiratory);there havebeen no systematicinvestigations of the ex- CT 06511 ), and Katherine S. Harris (Haskins Laboratories, 270 Crown tentto whichstress effects are manifest throughout the speech production Street,New Haven, CT 06511 and The GraduateSchool, City University system.The presentstudy investigates the effectsof stressin the vocal of New York, New York, NY 10021 ) tract and respiratorysystem. Lip and jaw and respiratorykinematics, laryngealvibration, and intra oral air pressurewere simultaneouslyre- While dynamicpalatography, a methodof measuringtongue-palate cordedduring recitation of sentencesin which lexical,contrasfive, or sen- contactover time, has a long history,the methodhas seenlimited use tential (focus)stress were varied. Changes in the articulatorykinematics becauseof the needto custom-makea palate for eachsubject. Recently, and in the relative timing of laryngeal,respiratory, and supralaryngeal the RionCompany has developed a "universal"semi flexible palate, with activityare analyzedas a functionof stresstype and positionin the utter- 64 contactpoints, in severalsizes. This methodologicalstudy was de- ance.Results suggest that the articulatorycorrelates of stressare distrib- signedto assessthe comparability of data from flexibleand custom-made uted continuouslywithin and acrosssubsystems. [Work supportedby palates,and the stabilityand reliability of flexiblepalate measures. Three NIH.] speakersrepeated multiple tokens of/isi/and /asa/with both palate types,and with flexiblepalates removed and replaced.While test-retest reliabilitywas satisfactoryfor flexiblepalates, the custom-flexiblecom- parisonsrevealed the needto developmeasures of subject-to-subjectdif- ferencesin palatesize and shape, if thetechnique is to developits potential for large-scalestudies of articulationin normaland clinicalpopulations. 10:57 Commentswill be made on analysissoftware, sampling rate, and contact- point distribution.[ Work supportedby NIH Grant NS-13617.] TTI2. Analysisof deaf speakers'breath streamdynamics during speech production. James J. Mahshie and Pradeep K. Yadav (Speech CommunicationLaboratory, Department of Audiology and Speech- LanguagePathology, Gallaudet University,Washington, DC 20002) Computersimulations have beenused to characterizeand better un- derstandthe articulatorybehavior of normal [M. Rothenberg,Biblio. Phon. ( 1968); E. Muller and S. Brown,Speech Lang. (1980) ] and disor- dered iV. Gracco and E. Muller, ASHA 13 (1981)] speakers.In the TT15. An evaluation of an alternating magnetic field device for presentwork, detailsof a computersimulation of breathstream dynamics monitoringarticulator movements. Shuyong Shao, Betty Tuller, and J. of speechproduced by deafspeakers will be presented.Oral airflow,oral A. Scott Kelso (Program in ComplexSystems and Brain Sciences, air pressure,nasal airflow, electroglottograph,and acousticdata were Florida Atlantic University,Boca Raton, FL 33431) acquiredfrom both deaf and normal speakersduring production of inter- vocalicbilabial plosives. Measures from thesedata were usedto establish The understandingof articulator11:33coordination during speech has been input parametersfor simulationof both the deaf and normal-hearing limitedby the type of datathat can be safely gathered. Specifically, nonin- speakers'utterances. Among the parametersmanipulated in the model vasivemethods of monitoringtongue motions are still notwidely avail- were:airflow resistanceat the articulatoryconstriction, nasal tract resis- able.One promising method is an alternatingmagnetic field device to tance, vocal tract enlargement,and glottal resistance.Pressure and flow track movementsof multiple articulatorsin the midsagittalplane. This outputsfrom both the simulationsand speakerswere compared. Advan- reportconcerns one such commercially available device, the Articulo- tagesand limitationsof the useof computermodels to understandspeech graphAG-100. The principlesof systemdesign and operation have been productionprocesses of disorderedspeakers will be discussed.[Work reportedpreviously by P. Schonleand colleagues, in Germany,who de- supportedby the Whitaker Foundation.] velopedthe system'sprototype. Here, an intensiveevaluation of system Sl 15 J. Acoust.Soc. Am. Suppl.1, Vol, 86, Fall 1989 118th Meeting:Acoustical Society of America Sl 15 accuracy,repeatability, and linearity, as well as measurements of tongue and hasto be trainedto the articulatorydata for the particularspeaker. positioningduring speech will be presented.These results indicate that [Work supportedby Ohio StateUniversity Speech and HeatingDepart- withproper precautions in transducercalibration and placement, the Ar- ment.] ticulographcan b• a usefultool in speechproduction research. [Work supportedby NIH andNIMH. ] 11:57 ß r17. Quantitative comparisonof articalatory models. Michel T. T..lackson (Divisionof Speechdnd HearingScience, Ohio State University,Columbus, OH 43210) 11:45 Phonologicaltheories invoke a numberof differentarticulatory pa- •16. Automaticalignment of phoneticevents with x-ray microbeam rametrizations.Traditional phonetic theories and some phonological the- articulatorydata and the acousticsl•ech signal.J. H. Greenwald,A. oriesparametrize vowel articulation in termsof tongucheight and back- K. Krishnamurthy, and O. Fujimura (Department of Electrical hesS.Particle phonology [ S. A. Schane,Phonology Yearbook 1, 129-155 Engineering,The OhioState University, 205 DreeseLabs., 2015 Neil (1984)] parametrizesvowel articulationas blendsof particular (ex- Avenue, Columbus, OH 43210) treme) articulations(i,u,a). Harshman,Ladefoged, and Goldstein Acoust.Soc. Am. 62, 693-707 (1977) ] parametrizeEnglish vowels using Thispaper describes an algorithm for the automatic alignment of pho- front-raisingand back-raising factors, each of whichgenerates a family of neticevents with x-ray microbeamarticulation data and the correspond- tonguepositions. A multispeaker,cross-linguistic database of x-ray trac- ingacoustic signal. The algorithmuses a two-stepprocedure similar to ings [M.T.T. Jackson,UCLA WorkingPapers in Phonetics71, 1-232 thatof Nelson[ Nelsonet al., J. Acoust.Soc. Am. Suppl.1 63, S32 (1978); (1988) ] allowsquantitative comparison of theseproposals. Each propos- Nelson,AT&T Bell LaboratoriesInternal Rep. (1978)]. The first step al definesa vowelspace: The coordinatesof vowelsin the vowelspace can locatesthe phraseboundary in continuousspeech, and the secondstep be fixed,and the articulatorycorrelates of the parametersusing multivar- matchesthe phonetic segments in eachphrase. Articulatory and acoustic iate regressioncan beestimated. Each model can, therefore, be evaluated eventsare recognized in continuousspeech, and matched to thepredicted bystandard methods, e.g.,)C a tests. A numberof such articulatory models phoneticevents using a dynamicprogramming technique. The placeof havebeen investigated. The resultssuggest that front-raisingand back- articulationand voicing for c•rtainphoneroes are also matched with ar- raisingfactors generalize well acrosslanguages and fit articulatorydata ticulatoryand acoustic events. The system is presently speaker dependent, betterthan height/backness-or particle-basedmodels. FRIDAY MORNING, 1 DECEMBER 1989 ST. LOUIS BALLROOM C, 8:30 TO 11:50 A.M. SessionUU. Underwater AcousticsVII: General Topics Robert Hickling, Chairman National Centerfor PhysicalAcoustics, Unioersity, Mississippi 38677 Chairman's Introduction---8:30 Contributed Papers 8:35 UUI. Abstract withdrawn. 8:50 UU2. Optimum bearing resolution for a moving towed array and extensionof its physicalaperture. Stergios Stergiopoulos (SACLANT UnderseaResearch Centre, Viale San Bartolomeo400, 19026 La Spezia, ltaly) This paperexamines the limits of the angular resolutioncapability of a moving towed array (MTA) by finding the Cramer-Rao lower bounds (CRLB) andproviding algorithms that extendthe physical aperture of an MTA. The model that is considered for the CRLB estimates assumes that an M-hydrophonetowed array is movingwith knownconstant speed and that in the receivedsignal unknowns are all the parametersfor two Sl 16 J. Acoust.Soc. Am. Suppl.1, VOL86, Fall 1989 118th Meeting:Acoustical Society of America S116 sources.The estimated CRLBs for this model indicated that an M-hydro- The theoreticalbounds have been validated by computersimulation and phoneMTA providesvery high angular resolution when the period Tof the simulationresults for a least-squaresestimator for variousarray/tar- thereceived signal is very long. This ability of the moving array to resolve getgeometries are presented. Methods by whicha passivesynthetic aper- twoclosely spaced sources isrelated to thefact that the physical aperture turemight be implemented are discussed and the results of a least-squares hasbeen extended by the distancetraveled by the arrayduring the T estimatoroperating on 120s of data recordedfrom a 10-elementtowed secondsof theobservation period. ½ nputersimulation examples using arrayare presented.[Work supportedby SERC and PlesseyNaval Sys- the maximumlikelihood estimator • ,d an extendedtowed array algo- tems.] rithm (ETAM) to find the bearingof the sourcesare presented.The resultsof thesesimulations agree with the CRLB if the signal-to-noise ratio (SNR) ishigher than 0 dBat thehydrophone level of-8 dB re: l-Hz 9:50 band.Real data applicationsusing the ETAM were successful,and the physicalaperture of a 32-hydrophoneMTA wasextended to an equiva- UU6. Acousticnavigatlonforalargeaperturearray. B.J. Sotirin (P-004, lentof a 512-hydrophonefully populated array during 185 s of observa- Marine PhysicalLaboratory of the ScrippsInstitution of Oceanography, tionperiod. These results have also indicated that the performance of the UCSD, La Jolla, CA 92093) abovealgorithm is very robust since it extendsthe physical aperture of an arrayby morethan one order of magnitudefor thecase of a verylow SNR The acousticnavigation system for a largeaperture (900 m) vertical broadbandsignal even though the source had 3.6-kn speed along its bear- array and the resultingdata, describingthe motionof the array and re- ing. searchplatform FLIP duringa 20-daydeployment in the NE Pacific,are presented.Acoustic navigation data are acquiredby detectingthe high- frequencysignals emitted by near bottomtransponders. Recorded travel timesare converted to xyzpositions by a nonlinearleast-squares approach 9:05 that adjuststhe transponder,FLIP, and the array positions,minimizing the rms error. Simulationresults highlight error sensitivities.The move- UU3. A new method for passivesynthetic aperture array processing. ment of FLIP, althoughconstrained by a three-pointmoor, exhibitsa EdmundJ. Sullivan(Naval UnderwaterSystems Center, Newport, RI wind-drivencomponent and a clockwisesemidiurnal tidal component 02841) and StergiosStergiopoulos (SACLANT UnderseaResearch with horizontaldisplacements in excessof 300 m. Suspendedvertically Center, APO New York, NY 09019) from FLIP under tension,the array hasa dampedresponse to tidal and internalwave forces. [Work supportedby ONT.] A techniquefor extendingthe usefulaperture of a towedarray is presented.Its advantageover previoustechniques is that it usesthe exist- ing apertureof the physicalarray to estimatethe phasecorrection factor for the successivemeasurement. Thus the need for an accurateknowledge 10:05 of the sourcefrequency is avoided.By usingthe fact that, at any time, the array overlapsits positionat a previoustime, a maximumlikelihood esti- UU7. Time domainsolutions for scatteringfrom objectsin a waveguide. matorcan be constructed that computesthe phasedifference between the Paul Ellis (Schoolof Physicsand Astronomy,University of Minnesota, measurementstaken at thesetwo positions.This estimatedphase differ- Minneapolis,MN 55455) and M. F. Werby (Schoolof Physicsand enccis then usedto correctthe secondmeasurement. An iterativeproce- Astronomy, University of Minnesota, Minneapolis, MN 55455 and dureis then usedto extendthis technique ore{ severalmeasurements. NORDA, NumericalModeling, Stennis Space Center, MS 39529) This estimationprocedure, along with estimatingthe kinematicphase Calculationsthat describe scattering from an objectin a waveguidein correctionfactor a•, includesany other phase errors that maybe incurred by sucheffects as array motion,propagation effects, or lossof temporal the frequencydomain are verytime consuming. For the timedomain it coherence.Results with bothreal and synthetic data arc shownthat dem- canbe orders of magnitudemore expensive both in timeand storage than onstratethat the techniquesignificantly outperforms previous tech- for thefrequency domain, rendering such calculations impractical unless niques. a supercomputercan be used. The solution of scattering from an object for a realistictarget and waveguideusing the Cray 2 computer(a 528- Megaword,4.1-ns machine) at theMinnesota Supercomputer Institute is discussed.The methodsemployed as well as resultswill be presented. 9:20 UU4. Inferring the shapeof a towedarray by processingthe data from its acoustic sensors.Brian G. Ferguson (Maritime SystemsDivision, 10:20 WSRL, DSTO, P.O. Box 706, Darlinghurst2010, Australia) UUS. Scatteringand transmissionof soundpower flow by solidelastic When a horizontallinear array of underwateracoustic sensors is spheresin water.Robert Hickling, R. Kirk Burrows,James F. Ball, and towedthrough the sea,motions within the water massand variationsin Lou Ann Redmond (National Center for Physical Acoustics, thetow vessel'sheading can cause the shape of thearray to deviatefrom a University,MS 38677) straightline. Distortionof the array's linear configurationcan lead to The scatteringand transmissionof soundpower flow, or soundinten- degradationin the detectionand localizationperformance of the array, sity, are calculatedfor planewaves incident on a solidelastic sphere in unlessthe spatialprocessing of the acousticdata can compensatefor the water.Since sound intensity is the samevector quantity in boththe solid nonlinearshape of thearray. One method for monitoringthe shape of the and the liquid media, the fiow of soundpower around and throughthe arrayinvolves instrumenting the array with compassesand depth sensors. spherecan be treatedas an entity. In additionto the flow of soundpower, Another techniqueis presentedhere wherebythe data from the acoustic distributionsof reactiveintensity and potentialand kineticenergy densi- sensorsthemselves are usedto infer the spatialdistribution of the sensors tiesare computed.Extensive use was made of Mathematicain this work. alongthe array. [Work supportedby the Officeof Naval Research.] 9:35 10:35 UUS. Performancebounds for passivesynthetic aperture sonar. Richard UU9. Farfield projection from sparse nearfield measurementsamples. William Harvey (School of Information Systems,University of East Giorgio V. Borgiotti (Department of Electrical Engineering and Anglia, Norwich NR4 7TJ, United Kingdom) Computer Science,George WashingtonUniversity, Washington, DC 20052) and Eric Rosen (Naval Kesearch Laboratory, Code 5130, The performancebounds for a towed line array are derivedand com- Washington,DC 20375-5000) paredwith performanceachievable using a staticarray. It is shownthat the towedarray can provideconsiderably better performancethan the The nearfieldto farfieldprojection procedure here introduced has the staticarray when the target is in the nearfieldof the syntheticaperture. followingnovel features. ( 1) The measurednearfield samples do notneed S117 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America Sl 17 to belongto a separablecoordinate surface. (2) The numberof samples curvesshows that the simplebicorrelation detector does not do nearlyas canbe as small as the theoreticalnumber of degreesof freedomof the far- wellas correlation or tricorrelationdetectors for transientswith 2rs >ft. field, relatedto the dimensionsof the sourcein wavelengths.(3) The Other performanceanalysis results are alsodiscussed. [Work supported samplesdo not needto belongto a regulargrid. (4) If the projectionis by NOARL and Navy/ASEE Summer Faculty FellowshipProgram.] soughtonly within a limited angularsector, a further reductionof the •) Also at NOARL. numberof measurementsamples is possible. The techniqueis based on the identificationfrom the measureddata of an equivalentmonopolar distri- butionof acousticsources on a surfacecompletely surrounding the radia- tor. The farfieldis then straightforwardly computed. The identificationis basedon the pseudoinversionof the matrixrepresenting the mappingof theequivalent sources into nearfieldsamples, which is obtainedby resort- ing to the singularvalue decomposition of the matrix. 11:20 UU12. Effect of pressureon soundabsorption in seawaterat 0 *C. F. H. Fisher (University of California, San Diego, Marine Physical Laboratory,Scripps Institution of Oceanography,San Diego, CA 92152) and W. M. Wright (Department of Physics, Kalamazoo College, 10:50 Kalamazoo, MI 49007) UUI0. Impact of samplewindow choice and location on time delay Earlier measurementsof theeffect of pressureon soundabsorption by estimation of transients. Allen E. Leybourne, III (University of magnesiumsolfate in seawaterby Hsu at 0 'C in the laboratorywere SouthernMississippi, Box 5137, Hattiesburg,MS 39406) and Roger limitedto pressuresof just over300 bars. The pressurewas limited by the W. Meredith (Naval OceanResearch and Development Activity, Arctic strengthof the 100-litertitanium sphere,a spareballast tank from the AcousticsBranch, Stennis Space Center, MS 39529) WoodsHole submarineALVIN. To obtaindata at higherpressures at 0 'C, Estimationof timedelays between coherent transient signals is biased measurementswere recentlymade in 0.5 M magnesiumsulfate solutions by thechoice and location of thesampling window. For transientsthat do in a cylindricalreesonator. Because the magnesiumsulfate relaxation fre- not tendto zerowithin thesampling window (self-windowing), the uni- quencyis independentof pressure,measurements are necessaryat only formor boxcarwindow may not be the best choice due to spectral leakage. onefrequency. This wasdone in the 100-kHzregion and a sodiumsulfate This is especiallyimportant when processing highly asymmetrical tran- solutionwas used to correctfor wall lossesin the cylinder.Mode interfer- sientsignals that are not self-windowing as in thecase of acousticambient enceproblems aswell as a shutdown inthe buildir•g air circulation system, noise.When a samplewindow is usedto suppressspectral leakage, a bias whichgave risc to temperaturecontrol problems, degraded the data. Nev- in timedelay estimation results from a suboptimalwindow location. Even ertheless,enough data wereobtained at higherpressures (600 to 1300 thoughthe boxcar window produces the least bias in timedelay when the bars) to be usefulin assessingthe effectof pressureon seawatersound transientis self-windowing, it is shown that sidelobe interference issignifi- absorptionat 0 'C ashas been done with earlierwork by Fisherat 25 'C on cant for highlyasymmetric signals. This work criticallyexamines these 0.5 molarmagnesium sulfate solutions. These results will beincorporated effectsand describes an iterativeprocedure to determine optimal location with thoseof Hsu at lowerpressures into an equationfor absorptionin of thetransient within any sample window. [Work supportedby theOf- seawaterat 0 øCcovering the whole pressure range. Predictions using this riceof Naval Technology.] equationwill be comparedwith thosefrom the Garrison-Francoisand Fisher-Simmonsequations for soundabsorption in seawater.[This work wassupported by Officeof Naval Research.] 11:05 UUll. Higher-order correlations and spectra and the detection of narrow-banddeterministic transients. Lisa A. Pfiug (NOARL, Stennis SpaceCenter, MS 39529and Departmentof Mathematics,University of 11:35 New Orleans,New Orleans,LA 70148), Kenneth H. Barnes,George E. Ioup, JulietteW. Ioup (Departmentof Physics,University of New UUI3. Geoaeoustiemodels for bottom loss predictions.Charles Orleans,New Orleans,LA70148),RobertL. Field, JamesH. Leclere,•) W. Holland (PlanningSystems, Inc., 7925 WestparkDrive, McLean, and GraysonH. Rayborn (Departmentof Physicsand Astronomy, VA 22102) Universityof SouthernMississippi, Hattiesburg, MS 39401) A useful theoretical measure of acoustic interaction with the seafloor Previousinvestigators have evaluated higher-order correlations and is the planewave reflection coefficient (R) whichis commonlyexpressed spectrafor detectionand time delayestimation of statisticalsignals. In asbottom loss (20 log [R I). Predictionsmodeling the seaflooras a sedi- this study detectionof known and unknown deterministictransients, mentlayer over a half-spaceindicate that bottomloss should be p.ropor- especiallythose which are narrowband, is considered.Initial resultshave tionalto frequency.Bottom loss measurements ( 50-1600 Hz) in thedeep beenreported [G. E. Ioup, J. W. Ioup, K. H. Barnes,R. L. Field,J. H. ocean,however, frequently show a lossthat is inverselyproportional to Ledere, and G. H. Rayborn, IEEE Workshopon Higher-OrderSpectral frequency.The locationsof theseanomalous sites are remarkablyconsis- AnalysisProc., 46-55 (1989) ]. For narrow-bandsignals extending from tentwith theboundaries of the AbyssalPlains which are characterized by frequencyf•tofr, the bispectrumis zerounless 2fs Sl 18 J. Acoust.Sec. Am. Suppl.1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America Sl 18 FRIDAY MORNING, 1 DECEMBER 1989 ST. LOUIS BALLROOM G, 9:15 A.M. TO 12:00 NOON SessionVV. PhysicalAcoustics IX: Propagationand PorousMaterial James M. Sabatier, Chairman NationalCenter for PhysicalAcoustics, P.O. Box 847, University,Mississippi 38677 ContributedPapers 9:15 sourceor an ultrasonictransducer. The acousticpressure in the borehole was measuredby a commercialPZT hydrophoneand by a home-made VV1. Studyof acoustictransmission loss in sedimentsat low frequencies PVDF hydrophone.The effectof the boreholeon the elasticwaves is using parametric acoustic arrays. V. N. Bindal, T. K. Saksena, S. identifiedby comparingthe measuredacoustic responses with the undis- K. Jain, Recta Gupta, and D. N. Santoshi (Ultrasonic Section, tortedincident pulse, calculated by thegeneralized ray theorywithout the National PhysicalLaboratory, New Delhi- 12, India) borehole.The theory of plane wave scatteringby a boreholeis usedto Investigationshave been carried out on the acoustictransmission loss comparewith the experimentaldata. The transientresults from boththe in clay, sand,and gravelin the frequencyrange 6-10 kHz. Parametric experimentsand the borehole theory agree with eachother very well, and acousticarrays have been used for this studyand havebeen generated they both exhibit resonancesdue to the multiple reflectionsinside the usingamplitude modulation of a primarytransducer of frequencyabout borehole fluid. 300kHz in whichthe carrier was suppressed. Source level and beam width of theparametric signal have been • 155dB re: I/zPa/m; and 3.6 ø, respec- tively. Experimentshave bccn conducted in the laboratorytaking the sedimentin a rectangularcontainer of dimensions0.7 X0.49X0.6 m3. The acoustictransmission loss using parametric arrays has been com- 10:00 paredwith the measuredloss with a nonparametricsource. The results VV4. Apparatusto determinethe complexmass density of a viscousfluid showa dependenceof transmissionloss on the particlesize and arc dis- containedin a rigid poroussolid from acousticpressure measurements. cussedin the light of previousstudies. Robert A. Mirick,'• Steven R. Baker, and Oscar B. Wilson (Department of Physics, Code 61Ba, Naval PostgraduateSchool, Monterey, CA 93943) 9:30 Two experimentaltechniques to determinethe frequency-dependent complexmass density of a viscousfluid contained in a rigid poroussolid VV2. Experimentaland theoretical wave propagation in solidand hollow are beinginvestigated. The fluid-filledsolid is containedwithin a small cylinders in fluid. Thomas J. Plona, SergioKostek, and Shu- (,(2) cylindricalcavity. In onetechnique the movingmass of the fluidis Kong Chang (Schlumberger-DollResearch, Old Quarry Road, sensedby the effectit hason the measuredinput electricalimpedance of a Ridgefield,CT 06877-4108) movingcoil transducer.In the secondtechnique the movingmass is ex- tractedfrom the measuredpressure required for the fluid to oscillatewith A comparisonbetween experiment and theory is made involving axi- a known amplitudethrough the poroussolid. Preliminary results have symmetricwave propagation along solid and hollow cylinders immersed beenreported for the impedancemethod [Grant et al., J. Acoust.Soc. in a fluid (i.e., water). The theoreticalmethods to generatewaveforms Am. Suppl.1 84, S176 ( 1988) ]. Preliminaryresults for thepressure tech- includeboth integral representation and finite-difference modeling. The niqueand a comparisonof the two methodswill be presented.[Work dispersivemodal characteristics are calculatedfrom the waveforms.The sponsoredby NRL-USRD and the Naval Postgraduate School.] experimentsare madeon solidand hollow steel cylinders of 0.75 in. and a•LCDR, USN. 0.625in. The sourceand receiver transducers are made from PZT rings 1.0 in. and 0.5 in. highwhich freely slide along the cylinders.Data are collectedas a functionof source-receiverspacing and over a frequency rangeof 40-240 kHz. A Prony'smethod is usedto obtain the modal characteristics.Discussions will focuson the goodagreement between theoryand experiment,the differencesbetween the modestructures in the 10:15 two casesand the asymptoticbehaviors at shortand longwavelengths. VVS. Measurementof acousticaland microstructuralproperties of real Someinteresting comparisons between moodpole and dipole modes will andmodel porous materials. Michael R. Stinson,Yvan Champoux,and also be made. Gilles A. Daigle (Division of Physics, National Research Council, Ottawa, Ontario K IA 0R6, Canada) Detailed measurementsof severalparameters describing porous mate- rials are currently being made. The acousticalbehavior of the porous 9:45 materialsis determinedusing an impedancetube technique, modified to VV3. Transientresponse of a fluid-filledborehole due to the passageof provideboth characteristicimpedance and propagationconstant, as a transientelastic waves. C. Changand S. K. Chang(Schlumberger-Doll functionof frequency.To definethe microstructuralproperties, flow re- Research,Old Quarry Road, Ridgefield,CT 06877-4108) sistivityand porosityare initially beingmeasured. Several real materials are beingconsidered, including samples of reticulatedopen-pore foam, The interaction of transient elastic waves with a fluid-filled borehole is felt materials,and naturalsoils. A seriesof modelporous materials, with investigatedboth experimentally and theoretically.The elasticwaves inci- well-definedmicrostructure, is also being developed.The first in this se- dentfrom variousangles are partiallyscattered by the borehole,while an riesconsists, simply, of circularpores of uniformcross section, formed in a acousticwave field is introducedin theborehole fluid. Experiments were rigid host material.The measurementswill be interpretedin terms of conducted in a solid block with a circular borehole inside. The transient existingrigid-frame models of theacoustical properties of porousmateri- elasticwaves were generated by usingeither a simulatedacoustic emission als. S119 J. Acoust.Soc. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America S119 10:30 11:15 VV6. The slow compressionalwave in a porousmedium by the finite VV9. Low-frequency, long-range sound propagation modeling over a element method.Yu-chiung Teng (Aidridge Laboratory of Applied locally reacting boundary with the parabolic approximation. J. Geophysics,Columbia University, New York, NY 10027) S. Robertson(Department of Mathematics,U.S. M. A., WestPoint, NY 10996-1786), M. J. Jacobson,and W. L. Siegmann(Department of The finite elementresults for a fluid/porous-solid/fluidsystem are Mathematical Sciences,Rensselaer Polytechnic Institute, Troy, NY presentedby usingthe double-nodel-layertechnique. The pressure-type 12180-3590) fluidelements are used to modelthe fluidfields: The poroussolid is mod- eledby Biot's medium [J. Acoust.Soc. Am. 28, 168 (1956) ]. The Biot's Thereis substantial interest in theanalytical and numerical modeling theorypredicts that for propagationof acousticwaves in fluid-saturated of low-frequency,long-range atmospheric acoustic propagation. Ray- elasticporous media, a secondcompressional wave exists. Geertsma and basedmodels, because of their frequencylimitations, do not alwaysgive Smit [Geophysics41, 169 (1961) ], usingBiot's theory, predicted the an adequateprediction of quantitiessuch as soundpressure or intensity generationof thissecond type of compressionalwave at anopen interface levels.However, the parabolicapproximation method, widely usedin betweena liquid and a poroussolid. Plona [Appl. Phys.Lett. 36, 259 oceanacoustics, and often more accurate than ray modelsfor frequencies (1980)] observedthe secondtype of compressionalwave in his liquid/ of interest,can be appliedto acousticpropagation in the atmosphere. porous-solid/liquidexperiment at ultrasonicfrequencies. Given suitable Modificationsof an existingimplicit finite-difference implementation for physicalparameters in the finiteelement calculations, excellent agree- computingsolutions to the parabolicapproximation are discussed.A lo- ment with the Plona's measurement has been obtained. The finite element cally reactingboundary is usedtogether with a one-parameter(the flow resultsfor a sealedinterface, showing a diminishingof thesecond type resistivity)ground impedance model. Intensity calculations are per- compressionalwave, also agree with the experimental work performed by formedfor a numberof flow resistivity values in bothquiescent and windy Rasolofosaon[Appl. Phys.Lett. 52, 780 (1988)]. atmosphericsound channels. Variations in thevalue of thisparameter are shownto havesubstantial effects on the spatialvariation of the acoustic signal.[Work supportedby NASA.] 10:45 VV7. Dependenceof the acousticto seismiccoupling ratio on theangle of 11:30 incidenceand geophonedepth. W. Pat Arnott, JamesM. Sabatier,and JohnO. Messer(National Centerfor PhysicalAcoustics, P.O. Box 847, VVI0. An investigationof the relationshipbetween upward refraction University, MS 38677) abovea compleximpedance plane and the sphericalwave evaluation for a homogeneousatmosphere. Richard Rasper, Gordon E. Baird, and An atmosphericsound wave can couple with the poroelasticground, WenliangWu (Physical AcousticsResearch Group, University of resultingin ground (or seismic)motion. This is the phenomenonof Mississippi,University, MS 38677) acousticto seismiccoupling. Microphones are used to measurethe sound pressurelevel at thesurface and geophones are used to measurethe result- A residueseries solution based on Fock'swork in electromagnetic ing seismicmotion at or belowthe surface.The seismic:acoustictransfer propagationhas beenused by severalinvestigators to examinesound function(SATF) characterizesa particular site. SATF measurementsare propagationinto a shadowzone. In thispaper it isdemonstrated that this reportedas a functionof the angleof incidenceof the soundwave for solutionmerges smoothly into the Sommerfeldsolution for soundpropa- frequencies15 Hz to 1 kHz. The angleof incidencevaried from 50 ø to 80ø, gationabove a fiat surfaceas the soundvelocity gradient goes to zero.A and normalincidence. Surface vertical and radial-horizontalgeophones principlethrust of thisinvestigation is thebehavior of thesound propaga- wereused. In additionsix vertical geophones at depthintervals of 10cm, tion abovea compleximpedance plane as the gradientbecomes finite. startingat 10 cm, were used.A seismicp-wave survey indicated a first Initial work indicatesthat the surfacewave pole may contributeto the layerdepth of 44 cmhaving a wavevelocity of 159m/s overlyinga layerof residue series solution under certain conditions. velocity379 m/s. This givesa critical angleof incidenceof 65•. It was anticipatedand confirmed that the SATF wouldincrease in magnitude for certainfrequency bands as the angleof incidenceapproached and passedover the critical angle.Comparisons of experimentalresults and multilayeredelastic and poroelasticmedia theory will be addressedas 11:45 timepermits. [Work supportedby ONR.] VVll. Scattering of sound by atmospheric turbulence. Walton E. McBride, Henry E. Bass, Richard Respet (Physical Acoustics ResearchGroup, The Universityof Mississippi,University, MS 38677), and KennethE. Gilbert (National Center for PhysicalAcoustics, The Universityof Mississippi,University, MS 38677) 11:00 A computersimulation of the effectof smallscale turbulence on atmo- VVS. Experimental studies of low-frequency propagation over fiat sphericsound propagation over a compleximpedance boundary is devel- terrain. D. Keith Wilson and Dennis W. Thomson(Department of oped.The atmosphereis brokenup into sphericallysymmetric eddies Meteorology,Pennsylvania State University, University Park, PA characterizedby a Gaussianprofile. Single scatter is assumed and a closed 16802) form of the firstBorn approximation for scatteringis obtained,giving eacheddie's contribution to thetotal fluctuation of thesound pressure at a Experimentalstudies of atmosphericacoustic propagation have in- receiverdownrange. The numericalsimulation was accomplished with cludedmonitoring the levelof a 27.7-Hz sourceat a distanceof 770 m. theconcept of a "realization,"or snapshotof the turbulentmedium. Each Two of theextended duration experiments consisted of recordingSPL at eddie'sscatter contribution was added up for a particularconfiguration of 1-minintervals for 75 h overdry ground(Oct. 1988),and for 144h over eddies,giving that realization'stotal soundpressure fluctuation. The ed- frozen,snow-covered ground (March 1989}. Comprehensivesurface lay- dieswere then givena randomchange in their coordinates.The total er micrometeorologicalmeasurements, including monitoring the turbu- soundpressure was calculated for thisrealization, and the process repeat- lent momentum,heat, and moisturefluxes adjacent to the path,as well as ed. A compleximpedance boundary was added and the predictionsof the continuousDoppler sodar measurements of the boundary layer wind pro- standarddeviations of the amplitudefluctuations, amplitude probability files, were made. These measurements are used to reconstruct the time- distributions,and structure functions were then tested against experimen- dependent"along path" soundspeed profiles. Interpretation of the acous- tal data.Good agreement was found whenever the average intensity of the tic dataincludes comparisons with theoreticalpredictions from a fastfield fluctuationswas well abovethe backgroundnoise level. [Work supported program (FFP). by U.S. Army ConstructionEngineering Research Laboratory. ] S120 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America S120 FRIDAY AFTERNOON, 1 DECEMBER 1989 ST. LOUIS BALLROOM B, 1:00 TO 4:15 P.M. SessionWW. PsychologicalAcoustics XI and Musical AcousticsIV: Perceptionof ComplexSound Charles S. Watson, Chairman Departmentof Speechand HearingScience, Indiana University,Bloomington, Indiana 47405 ContributedPapers worsewhen the changedsequence was transposed in frequency[G. R. Kidd andG. S. Watson,J. Acoust.Soc. Am. Suppl.! 84, S141(1988) ]. WWl. Weighting functions in a multiple tone task. Bruce G. Berg Listenersapparently rely on absolutefrequency whenever they can and (PsychologyDepartment, University of Florida,Gainesville, FL 32611) arerelatively poor at usingrelative-frequency information even when they An analyticaltechnique is usedto assessthe relativeweight given to mustdo soto performa discriminationtask (i.e., whenthe changedpat- eachobservation in a multipletone task.After listeningto a sampleof ternis transposed ). The abilityto detectchanges in relativefrequency was seventones presented sequentially, observers try to discriminateif the examinedfurther in newexperiments that includedboth high- and low- toneswere sampled from a "high" or "low" distribution.The frequencyof uncertaintyconditions with five-toneand two-tonesequences. Under eachtone is determinedby samplingfrom oneof two Gaussiandistribu- high-uncertaintyconditions, performance with bothfive- and two-tone tionswith meansof 1000and 1100Hz. Reliabilityof eachtone is manipu- sequenceswas again relatively poor when the patternswere transposed, latedby changingthe standard deviation of thedistributions as a function althoughsomewhat better with the two-tonesequences. Under low-un- of temporalposition. In experiment1, the threecentral tones have either certaintyconditions, performance with transposedpatterns at both se- equal,greater, or lessreliability than the remainingtones. Observers ad- quencelengths was often as goodor betterthan with untransposedpat- just their weightsappropriately. In experiment2, sequentiallypresented terns,suggesting a strong role of patternfamiliarity in theidentification of tonesalternate between high and low reliability,and intensitycues their patternsby the sequencesof relativefrequencies. [Work supportedby reliability.Observers adjust their weights appropriately when highly reli- AFOSR and NIH. ] abletones have greater intensity, but not whenhighly reliable tones are cuedby low intensities.Finally, in a two-toneexperiment, only the first toneof fixedintensity is reliable.Unexpectedly, discrimination accuracy decreasesas the intensityof the second,uninformative, tone is increased. [Work supportedby AFOSR. ] 1:45 WW4. Fatigueeffects of a puretone and puretones. I. M. Youngand L. D. Lowry(DepartmentofOtolaryngology, Jefferson MedicaI College of 1:15 ThomasJefferson Universily, Philadelphia,PA 19107) WW2. Effect of multiple presentationson detectionand recognitionof Subjectswith normal hearing were exposed to moderateintensity of a puretones and tonal patterns.Robert C. McKinley and Daniel L. Weber puretone and pure tones. Exposure frequency consisted of: a singletone, (Department of Psychology,Wright State University, Dayton, OH 1500 Hz; two tones,1500 and 2000 Hz; and threetones, 1500, 2000, and 45435) 3000 Hz. Fatigueeffect measurements were made by automaticaudi- ometryimmediately after exposure.The greatestfatigue at frequencies This studyexamines the effectof multipleobservations on detection I/2-1 octabove an exposurefrequency of 1500Hz wasshown by a single performance,and on the ability to recognizeone signalfrom a set of tone, the leasteffect by three tones,and the intermediateeffect by two possiblesignals, for threedifferent sets of foursignals. One set is fourpure tones.The fatigueeffects at the remotehigh frequencies were in the re- tones(100-ms sinusolds, frequencies of 707, I000, 1414,and 2000 Hz). verseorder. These findings may be explainedon the basisof harmonics Two setscontain tonal patternseach made of seven100-ms sinusoidal and combinationtones arising in the ear on stimulationwith tones. components.The firstset of fourpatterns used different frequencies from different one-fourth octave bands (centered at 707, i•00, 1414, and 2000 Hz) to producea singlepattern of relativefrequencies in differentfre- quencyregions. The secondset had differentsequences of a singleset of frequencies.All stimuliwere adjusted to equalsensation level in the pres- enceof a continuouswhite noise(spectrum level 20 dB SPL). Each trial 2.-OO had multiplepresentations (n = 1, 2, 4, 8, or 16) of a givensignal plus noise(probability of a signaltrial was 0.5) or noisealone. Detection WWS. Auditory-filter shapesat three test frequencieswith a burst and a performancefor the puretones and the firstset of tonalpatterns was well fringed notched-noisemasker. Beverly A. Wright (Department of describedby a squareroot of n improvement;detection for the secondset Psychology,University of Texas,Austin, TX 78712} of patternswas not. Recognition performance was generally poorer than The shapesof theauditory filters centered at 0.5, 1.0,and 3.0 kHz were that predictedby the detection/recognitiontheorem [S. J. Starr, C. E. determinedusing Patterson's notched-noise method, but with theonset of Metz, L. B. Lusted, and D. J. Goodenough,Radiology 116, 533-538 the tonalsignal delayed either I ms (burst condition),or 350 ms (fringed (1975) ] and was not affectedby multiple presentationsother than as condition)after the onset of themasking noise. The maskingnoise ranged predictedby the increasein detectionperformance. [Research supported from 0.1-10.0 kHz, and itsspectrum level was 24 dB. All waveformswere by AFOSR throughAAMRL/BBA, Wright--Patterson AFB. ] gatedwith a rise/fall time of 10 ms, and the offsetof the signalalways precededthe offsetof the noiseby 2 ms. Althoughthere were individual differencesacross listeners, for all six of them, the filters at the three test 1:30 frequencieswere narrower in the fringedthan in the burstcondition. In WW3. Detection of ratatire-frequencychanges in tonal sequences.Gary general,the differencesin the filter shapesin the two conditionsincreased R. Kidd and CharlesS. Watson (Department of Speechand Hearing as the signalfrequency increased. In accordwith unpublishedwork of R. Sciences,Indiana University,Bloomington, IN 47405) Carlyon,the narrowingprocess was complete with signaldelays of 50- 100ms and was observed for signaldurations of 25 and 50 ms,but not for a It waspreviously reported that listeners'abilities to detectfrequency signalduration of 200 ms. Possibleunderlying mechanisms will be dis- changesin randomly generatedsequences of toneswere substantially cussed.[Work supportedby NINCDS.] S121 J. Acoust.Soc. Am. Suppl. 1, VoL86, Fall 1989 118thMeeting: Acoustical Society of America S121 2:15 measureson thresholdand signallevel. Auditory-filter characteristics, criticalratios, psychophysical tuning curves, and masking patterns were WW6. Frequency resolution for broadband-noise masked normal obtainedfrom normal-hearinglisteners in quiet and in the presenceof listeners.Judy R. Dubno and Amy B. Schaefer(Division of Head and broadbandnoise at fivelevels ( 30 to 70dB SPL). Thesefindings were then Neck Surgery,31-24 RehabilitationCenter, UCLA Schoolof Medicine, comparedto previouslyobtained frequency-resolution results for subjects Los Angeles,CA 90024-1794) with sensorineuralhearing loss whose auditory thresholds in quietwere comparableto the normal-hearinglistener's masked thresholds. These Whenevaluating frequency resolution of hearing-impairedlisteners, comparisons,while limited to individualswith moderatehearing loss and it is often difficult to determineif changesobserved in auditory analysis broadband-noisemasked normal listeners, suggest that frequencyresolu- with increasingauditory threshold reflect abnormal cochlear function or tion observedin hearing-impairedlisteners is similarto that seenin the representnormal auditory processing at highstimulus levels. This experi- normal auditorysystem at high stimulussound-pressure levels. [Work mentwas designed to investigatethe dependenceof frequency-resolution supportedby NIH NS 18015.] 2:30-2:45 Break 2:45 3:15 WW7, Gap detectionand discriminationin complexspeechllke signals. WW9. Effectsof a silenttemporal interval on discriminability of burstsof PeggyB. Nelsonand SusanJ. Norton (Universityof KansasMedical reproduciblenoise. Susan M. Fallon (Department of Psychology, Center,Kansas City, KS 66103) Universityof Florida, Gainesville,FL 32611) and Donald E. Robinson (Department of Psychology,Indiana University, Bloomington,IN Speechlikestimuli consistingof/s/-like noiseand F 1-like frequency 47405 ) sweepswere usedfor gapdetection and discriminationexperiments. Five normal-hearingsubjects participated in the tasks:( 1) detectionof gapsin A same-differentpsychophysieal method was usedto examinethe /s/-like noise,(2) discriminationbetween gaps in noise,(3) detectionof discriminabilityof burstsof reproduciblenoise. Bursts in a pair wereiden- gapsbetween noise and glide,and (4) gapdiscrimination between noise tical on "same"trials. On "different"trials, bursts were identical except and glide. Stimuli for the latter two experimentswere patternedafter for •- msof independentnoise located at eitherthe beginning,middle, or synthesized"say" and "stay" speechsignals. These stimuli consisted of endof the burstpairs. Fallon and Robinson[J. Acoust.Soc. Am. Suppl. 1 120 ms of high-pass-filterdigital white noise(simulating the/s/pho- 78, S46 (1985) ] showedthat discriminabilitydepends on the temporal neme),followed by a silentgap varying from 0-120 msin duration,and a positionand proportionalduration of the independentnoise. The present 50-mslinear frequency sweep from 230-611 Hz (simulatingthe F 1 transi- experimentsexamined whether a silentinterval placed either immediately tion}. The samehigh-pass-filtered white noisewas usedfor all experi- prior to or followingthe independentnoise improves discriminability. ments.An adaptivepsychophysical procedure was used in the first two Resultsshowed that a silentinterval only improveddiscriminability if it experiments;a methodof constantstimuli procedurewas usedfor the replaceda portionof the noisecommon to the pairs of bursts.In this complexstimuli tasks because of the shallow-slopingpsychometric func- experiment,a briefburst of commonnoise temporally separated from the tions. Subjectsrequired extensive training to reach asymptoticperfor- independentnoise by 90-120 ms causeda large decrementin perfor- mancelevels on the latter two experimentaltasks. Results indicated sig- mance.These data suggestthat interactionsoccurring after sucha long nificantlylonger durations are requiredfor the complexstimulus tasks. silentinterval are unlikely to begenerated peripherally. [Work supported by AFOSR. ] 3:00 3:30 WW8. Gap detectionin narrow bandsof noiseas a functionof the number andproximity of the bands.J. H. Groseand J. W. Hall, III (Divisionof WW10. Auditory temporal aeuity for dynamic signals. Lawrence Otolaryngology/Headand Neck Surgery,University of North Carolina L. Feth, Mary E. Neill (Department of Speechand Hearing Science, at ChapelHill, ChapelHill, NC 27599-7070) Ohio State University, Columbus, OH 43210), and Ashok Krishnamurthy (Department of Electrical Engineering,Ohio The minimumdetectable gap in a noisedecreases both as the band- State University, Colombus,OH 43210) width of the noiseincreases and asthe numberof independentnoise bands carryingthe gap increases.The presentstudy attempted to relatethese Previousreports of auditorytemporal acuity have generally used stat- two observationsby differentiatingbetween the parametersof bandwidth ic, rather than dynamic, signals;that is, the charactersiticsof the signal and number of active channels. The stimuli consisted of either one band or did not changewithin its briefduration. Temporal acuity is definedas in fiveindependent bands of noise,each 20 Hz wide.In all conditions,one of Green [Psychol. Rev. 78, 540-551 (1971)]: the shortesttime interval the bandswas centered on 1000 Hz. The multiple bandswere eitherjuxta- within which a listenercannot detect a changein the sound.Auditory posedto givea singleband of noisecentered on 1000Hz or werespaced in temporal acuity is determinedusing frequency-modulated (FM) sinu- variousconfigurations, resulting in a set of conditionsranging from all solds.The two FM signalsused in eachblock of trial alwaysbegan at the bandspassing through a singleauditory channel to eachband passing sameinitial frequencyand movedto the samefinal frequency.Only the throughan independentchannel. The stimuliwere presentedat a spec- trajectory, the path from initial to final frequency,differed. One trajec- trum levelequal to that ofa 15-dBSL, 20-Hz-wideband of noisecentered tory, the glide, movedlinearly from initial to final freqency;the other on 1000Hz. The gapwas imposed with a 10-msrise-fall time. Gap thresh- proceededin a seriesof discretesteps. As the numberof stepsincreased, oldsimproved with an increasedcumulative bandwidth, but moreso for the sizeand durationof an individualstep decreased, and the multiple- the spacedbands than for thejuxtaposed bands. [Research supported by steppedtransition more closely approximated the glide.Four well-prac- AFOSR. ] ticedlisteners were asked to identifythe intervalcontaining the glidein a $122 J. Acoust.Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America $122 2-Q, 2AFC task.Frequency transitions of 100-,200-, and403-Hz dura- the sequences.A modelbased on computationof thesample correlation, tionsof 25,50, and 100 ms were centered on octave frequencies from 250 limitedby an internaltemporal variability of approximately15 ms,pre- to4 kHz. Signalswere presented at 50 dB SL foreach subject. Results for dictedobserver performance in a varietyof conditions.[Work supported the four listenerswere remarkably consistent, with the 75% discrimina- by AFOSR. ] tionpoint falling at 7 to 10ms for all conditionsexcept for signals centered at 4 kHz. Poorerperformance at 4 kHz mayreflect the lossof synchrony in the frequencyregion; however, there are somedifficulties with that explanation.[Work supportedby a grantfrom AFOSR.] 4:00 WWI2. Temporal DLs in isochronicloudness and pitch patterns. Caroline B. Monahan and Ira J. Hirsh (Central Institute for the Deaf, 818 South Euclid, St. Louis, MO 63110) In an adaptiveprocedure, listeners discriminated the delay of a single 3:45 tonein otherwiseisochronic six-tone patterns. The tern poral DL wasmea- WWII. Discriminationof partially correlatedtemporal sequences. R. suredat threetempos (isochronic intervals of 50, 100,and 200 ms) and D. Sorkin,M.V. Pezzo,and G. C. Elvers(Department of Psychology, for fourdelay positions (tones 2-5 ). In experiment1, patternscomprised Universityof Florida,Gainesville, FL 32611) loud (forte) and soft(piano) tones:thus FFPPPF, PFPFPF, etc.,where F was 80 dB SPL and P was 70 dB SPL. In experiment2, frequency How doobservers determine whether two equitonesequences have the patternswere "parallels" of the loudnesspatterns of the firststudy: thus sameor differenttemporal patterns? On each trial of this experiment, HHLLLH, LHLHLH, etc., whereH(igh) was 1667Hz and L(ow) was observerswere presented with two sequencesof n tones(n = 8, 12, or 16; 1000 Hz. In both studies,the measureof timing accuracy(Weber ra- tone duration= 35 ms; frequency= 10130Hz; level= 71 dB). The se- tio X 100) was similar to that found in earlier literature for the discrimina- quenceswere played successively and theobservers had to reportwhether tionof singletime intervals, namely, 5%-8% at 200ms, 9%-15% at 100 thetemporal pattern ofintertone gaps was the same or differentin thetwo ms,and 15%-25% at 50 ms.Accuracy was poorest for earlyserial posi- sequences.On half of thetrials, the sequencesof gapswere ideatical (per- tionsof thedelayed tone. This effect occurred primarily at 50 msand was fectlycorrelated, p = 1), and on half of the trials the sequenceswere significantonly in theloudness experiment. In experiment1, at 50 and 100 partiallycorrelated (0 FRIDAY AFTERNOON, 1 DECEMBER 1989 ST. LOUIS BALLROOM A, 1:1• TO 3:39 P.M. SessionXX. SpeechCommunication X: Acousticsof Speech Marios Fourakis, Chairman CentralInstitute for the Deaf, 818 SouthEuclid Avenue,St. Louis,Missouri 63110 ContributedPapers 1:15 diphthongs.Results of thisevaluation, in additionto durationalmeasure- meatsof individualdiphthongs and their componentsegments, will be XX1. Some acoustical properties of diphthongs.Michael Gottfried reported.[Work supportedby NINCDS.] (Central Institute for the Deaf, 818 South Euclid, St. Louis, MO 63110) Oneapproach to describingdiphthongs is to considerthem as consist- ing of a steady-statenucleus and a glidetoward another steady-state seg- ment. This studyevaluates this approachwhen sourcesof acousticvari- ation (change of speaker,stress, and rate) are introduced.In a 1:27 preliminarystudy, four speakersof MidwesternAmerican English were XX2. Acoustic characteristics of Xhosa dicks. Bonny E. Sands recordedproducing diphthongs in isolationand in the context/K_T/. (Departmentof Linguistics,UCLA, Los Angeles,CA 90024-1543) On the basisof the firstthree formants, diphthongs can indeed be charac- terizedin termsof two kindsof spectrallydefined states, steady-state and Xhosa,a Bantulanguage of SouthernAfrica, has 15 distinctclicks. glide.There is, however, variation in thenumber and order of thesestates Clicksare madeat threeplaces of articulation--dental,alveo-palatal. and for particulardiphthongs. Plots of thesespectral patterns in an auditory- lateral--and with five differentaccompanimeats---voieeless, aspirated, perceptualspace [J. D. Miller, J. Acoust.Soc. Am. 85, 2114-2134 nasalized,murmured, and nasalized murmured. A recordingwas made of (1989) ] suggestedthat diphthongscan be differentiatedby a combina- eightspeakers saying words containing each of these! 5 clicksbefore each tion of distinctivenucleus zones and angular movementswithin the space. of Ihe •ive vowels/i,e,a,o,u/. The DSP Sona-Graphwas usedto produce Productionsby two speakersof a carriersentence containing diphthongs widebandspectrograms, waveforms, and spectraof theseutterances. undervaried stress and rate conditionsare beinganalyzed. Results will be Dental clickswere marked by fricativenoise occurring throughout a wide usedto evaluatethe usefulnessof the descriptivescheme noted above and frequencyrange, and by lackof a clearonset. Alven-palatal clicks were theability of theauditory-perceptual space to capturedifferences between markedby an abruptrelease with little frication.Lateral clicks were more S123 J.Acoust. Soc. Am. Suppl. 1, Vol. 86, Fall 1989 118thMeeting: Acoustical Society of America Sl 23 affricatedthan the alveo-palatalclicks. The placeof articulationof the orderformants, third and above, when present for certain vowel sounds, click wasnot observedto haveany effecton the formanttransitions of the may be of particular importanceas diagnostictools in vocal-fold-vocal- followingvowel. Observed vowel effects on clickswere limited to the tract clinical examination. reductionin amplitudeof high-frequencynoise before rounded vowels /u/and/o/. The five accompanimentsat eachplace have a numberof phoneticcharacteristics that will be discussed.[Work supportedby NSF. ] 2:15 XX6. Automatic correctionof formant tracks: A progressreport. J. Andrew Fingerhutand James D. Miller (Central Institutefor the Deaf, 1:39 818 South Euclid, St, Louis, MO 63110) XX3. Statistical analysis with fundamental frequenciesof 12 Korean Sincecommercially available formant trackersare ofteninconsistent, monophthongs.H. J. Shin and S. W. Yoon• (AcousticsResearch muchhand-editing is requiredto correcttheir outputs.A softwarepack- Laboratory, Department of Physics,Sung Kyun Kwan University, ageis currentlybeing developed that may reducethis problem.Two fe- Suwon,440-746, Republic of Korea) male and two male speakerswere recordedproducing ten vowelsin S- vowel-T context, each vowel twice. Boundariesbetween burst-friction, The 12 Koreanmonophthongs pronounced by fivemale vocal musi- glottal-source,and silentsegments are locatedusing zero crossing and cians,each at fivefundamental frequencies, are studiedby a methodof rms energymeasurements. Pitch and formant-frequencyvalues are ex- statisticalanalysis with the fundamentalfrequencies (SAFF) basedon tractedusing the API and SGM commandsof ILS. Our softwarecorrects theformat analysis. In thisstudy, tbe SAFF methodshows more reasona- the pitchcontour and calculatesa sensoryreference from it [J. D. Miller, ble resultsof the format frequenciesin the Korean monophthongs.It is J. Acoust.Soc. Am. 88, 2114-2134 (1989) ]. Rulesbased on the relation alsoobserved that the first and secondformants are characterizedby the betweenpeak frequenciesand this sensoryreference are utilized to label resonanceof the cavitiesof thepharynx and mouth, respectively. The lip- the peaksas formants and the resultingformant tracks are then low-pass roundingeffect decreases the second-formantfrequency. a• Current ad- filtered.In this way, satisfactorytracking is obtainedfor all 80 syllables. dress:National Center for PhysicalAcoustics, University of Mississippi, [Work supportedby AFOSR and NIDCD. ] University,MS 38677. 2:27 1:51 XX7. Specificationof vowelheight using F 1 trajectories.H.S. Gopaland XX4. Effects of Consonantcontext and lexical stresson vowel formant Joyce Manzella (Department of Speech and Hearing Sciences, frequenciesin continuousspeech. Caroline B. Huang (Departmentof University of California at Santa Barbara, Santa Barbara, CA 93106) ElectricalEngineering and Computer Science and ResearchLaboratory of Electronics,MIT, Room36-511, Cambridge, MA 02139) Recently,the claimhas been made that both acousticand perceptual ambiguitiesbetween high and low vowelscan be reducedwhen F 1 time/ The effects of consonant context and lexical stress for vowels and frequencyvariations, rather than F 1 "target" values, are considered vowdsin continuousspeech have been investigated in previousacoustic [Di Beneditto,J. Acoust.Soc. Am. 86, 55-66 (1989) ]. The currentstudy studies[e.g., K. Stevensand A. House,J. SpeechHear. Res. 6, 111-128 investigatesthe efficacy of theF 1time/frequency variation ( theF I trajec- (1963); P. Dclattre, Final Rep., Universityof California (1969); F. tory) as an acousticmeasure of vowel heightusing an extendedset of Koopmans-vanBeinum, Ph.D. thesis, University of Amsterdam vowels.Additionally, the effectof speakingrate on thesetrajectories as it (1980) ]. However, the studiesof consonantcontext and lexical stress relatesto vowel height is investigated.Four high-low vowel pairs in involvedonly isolated words or wordsin a carrierphrase, and the studies /pVC/target syllablescontaining voiced and voicelessstops and frica- of continuousspeech did not considereach consonant context separately. tires are included.These syllables were produced at variousrates in sen- In the presentstudy, formant frequencies are measuredin vowelstaken tencecontext by four AmericanEnglish speakers, two malesand two froma readstory that is well controlled with respectto consonantcontext females.F I trajectoryversus F1 maximummeasures are discussedin and lexical stress.The vowelsstudied are/iY/,/eY/,/t/,/•/, and/^/. terms of their relativeefficacy in classifyingvowels into high and low The consonantcontexts are fo/, /d/, /g/, /w/, /r/, and/1/. Tbe vowels groups. carry primary or secondarylexical stress. The samevowels and contexts are elicitedin a retold story and in nonsensewords spoken in a carrier phrase.Preliminary results from one speaker indicate that consonant con- text and lexicalstress do affectvowels in continuousspeech, as expected. Furthermore,results so far showno gross differences in theeffects of these 2:39 factorson vowdsin continuousspeech and in the isolatedwords studied previously.[Work supportedin partby NIH.] XXS. Acousticcomparison of unvoicedfricatives in telephonespeech versustypical speech. David W. Petr, Todd G. Ellenberger,and Victor S. Frost (Telecommunicationsand Information SciencesLaboratory, Universityof Kansas,Lawrence, KS 66045) The resultsof a pilot studydesigned to comparethe acousticproper- 2:03 tiesof unvoicedfricative$ in telephonespeech versus speech typically used XXS. Higher harmonicsand formantsanalysis of speechsignals. Victor for acoustic-phoneticanalyses are reported. Two newcharacterizing pa- F. Coroneland Sultan Catto (Departmentof Natural Sciences,Baruch rametersfor the unvoicedfricatives are introduced: the half-power band- College, 17 LexingtonAvenue, New York, NY 10010) width (resonancewidth ) anda three-pointenergy versus time contour. It is foundthat the averagefricative power normalized by the averageadja- Spectralanalysis of windowedsegments ( 12.8-52.8 ms in duration), cent vowel poweris considerablylarger in the telephonespeech for all of theapproximately steady portion of vowelsounds, in thecontext CVC, speakers,fricatives, and fricativepositions studied. Also, half-power wasanalyzed for severalindividuals at variousSPL levelsfor eachsubject. bandwidthstend to be larger for [f] and [0] in telephonespeech, but It wasobserved that the signalswith larger SPL valueswere much richer smallerfor [s] and [1]. The studyyields a preliminaryunderstanding of in harmonicsand formants.A relationbetween the strainingof the vocal theseacoustic differences in terms of the differencesin vocal effort, micro- cordsand the higherharmonics is establishedand its relationwith nonlin- phoneplacement, and frequency response and bandwidth of therecording ear modelsof the vocal chordsand vocal tract examined.The higher equipment.Finally, possible distinguishing perceptual cues for thediffer- harmonicsof thefundamental frequency, sixth and above, and the higher- ent recordingenvironments in terms of the acousticdifferences are hy- $124 J. Acoust.Soc. Am. Suppl.1, Vol. 86, Fall 1989 118th Meeting:Acoustical Society of America $124 pothesized.In particular,the energy-time contour of [ 0] exhibitsa pecu- at frequenciesabove 2000 Hz. The implicationsof differencesbetween the liar high4ow-highpattern that is most pronouncedin the telephone spectrafor testingwill be discussed. recordings.[ Work supported by AT&T, NSF, andthe University of Kan- 3:15 2:51 XXll. Accuratemeasurements of jitter, shimmer,and glottai turbulence XX9. Perceptualjudgments of modifieddeaf speech.Grace H. ¾eni- noise from sustained vowels. Hideki Kasuya, Yoshinobo Kikuchi, Komshian(Department of Hearingand SpeechSciences, University of Zaoming Huang, Osamu Kernufo, and Keisoke Maeda {Faculty of Maryland, College Park, MD 20742), H. Timothy Bunnell (A. !. Engineering,Utsunomiya University, Utsunomiya 32 I, Japan) Dupont Institute, Wilmington, DE 19899), and James J. Mahshie (GallaudetUniversity, Washington, DC 20002) There hasbeen much research on the measurementof jitter, shimmer, and glottalturbulence noise from sustainedvowels, primarily aiming at The perceptualeffects of modifyingspeech produced by deaftalkers acousticquantification of pathologicalvoices. Many problems remain un- were investigatedto examine,in part, the changesnecessary for disor- solved,however, specifically those of accuratemethods for the pitchex- deredspeech to be judged normal. Recordingsof the rainbowpassage traction and noisemeasurement and of quantificationof shimmer.This producedby a numberof deafmale speakers were studied and threepas- paperdescribes a sophisticated method for thepitch period determination sagesjudged to bemoderately disordered were selected. The word "beau- basedon a cycle-to-cyclesimilarity of the waveformand a novelmethod tiful" wasextracted from thesepassages and from a similarpassage re- to quantifyshimmer characteristics. In thismethod, shimmer is analyzed cordedby a hearingtalker. Each of the deaf talkers'productions was for the individualharmonic component of the waveformin termsof its pairedwith the normalproduction to maketwo typesof continua.One variationsin amplitude and phase characteristics.Acoustic measure- type was basedon a time warpingof the disorderedspeech to fit the mentsof glottal turbulencenoise due to insufficientvocal-fold closure are temporalstructure of the normal speech,while the other was basedon a alsodiscussed. Efficacy of the newmethods for the measurementof jitter, time warpingof the normalspeech to the temporalstructure of the disor- shimmer,and noiseis comparedwith previouswork usingpathological deredspeech. Each continuum contained ten steps,ranging from disor- voicesamples. [Work supportedby a Grant-in-Aid from the Ministry of dered to normal spectralproperties. Graduate studentsin the fieldsof Education,Science and Culture,Japan. ] speechand hearingjudged whether the stimulisounded normal or disor- dered.Continua based on the normaltalker's timing providedordered responsesshowing an increasingproportion of "normal" responsesfor stepsapproaching normal spectra. For continuabased on thedeaf talkers' timing,however, responses were generally not orderedacross the spectral continuum.The relativemerits of temporaland spectralmodifications 3:27 will be discussed. XXI2. Soundtransmission to the chestwall in the subglottalrespiratory systemat low frequencies.George R. Wodicka (Schoolof Electrical Engineering,Purdue University, West Lafayette, IN 47907} The amplitudeof soundtransmission from the mouthto sitesoverly- 3:03 ing the extrathoracictrachea and on the posteriorchest wall wasmea- XX10. Power spectra of some modesof speech.Igor V. Nfib•iek suredfrom 100 to 600 Hz. Bandlimitedwhite noisewas input into the (Departmentof Audiologyand SpeechPathology, College of Liberal mouth of eachsubject with the glottisopen at restinglung volume,and Arts, Universityof Tennessee,Knoxville, TN 37996-0740) measurementswere performedusing aecelerometers. The ratiosof the transfer functions between the tracheal site and the source, and the chest The data to be reportedrepresent an extensionof the previousstudy wall siteand the source, were estimated. All of theresulting spectra exhib- "A comparisonof spectraof loudand whispered speech" [Nfib•lek and ited a peak that occurredfor the group at 129 4- 6 Hz with a Q of Maroonroge,J. Aeoust.Sec. Am. Suppl.1 75, S83 (1984) ]. Powerspec- 2.2 q- 0.4. Also, a trend of decreasingspectral energy as frequencyin- tra weredetermined for threemodes of runningspeech: phonated conver- creasedfrom the peaklocation was observed ( - 17, 4- 5 dB/oct). The sationalspeech, phonated monotonous speech, and whisperedspeech. fundamentalradial mode era modelthoracic cavity, which is a largerigid The averagespectrum of the phonatedconversational speech had a broad cylinderfilled uniformly with losslesslung tissue,provides a goodesti- maximum around 300 Hz; the spectraof the phonatedmonotonous mateof this low-frequencyresonance. This agreementsuggests that tho- speechhad pronounced peaks at the fundamentalfrequency and its har- raciccavity resonances may have particularly important effects on sound monics.The averagespectrum of the whisperedspeech was rather flat at transmissionin the subglottal•spiratory systemat frequenciesbelow low frequencies,and it wassimilar to thespectrum of thephonated speech roughly250 Hz. $125 J. Acoust.Sec. Am. Suppl. 1, Vol.86, Fall1989 118thMeeting: Acoustical Society of America $125