EFFECTS OF LISTENING EXPERIENCE ON DECODING SPEECH IN HKLIOX ENVIRONMENTS Harry Hollien, Ph.D. Carl L Thompson,Ph. D. IPCP andLinguistics Universityof Florida Gainesville,FLORIDA 32611 U. S A. Veteran'sAdministration Medical Center, NewOrleans, LOUISIANA 70146U.SA. Thepurposeofthis experiment wasto assess 1!the ability of auditors todecode speechproduced inthe He02 envimnment and2! the effects oflistening experience uponthis skill/ability. Three pai red groups ofauditors listened toequated speech tasksand were tested on their ability to decode the heard utterances. The samples wereproduced bydivers situated inan underwater habitat atdepths upto 1000fsw Ofeach pair, one listener group was subjected todaily exposure ofHe02 speech samples no feedback! for twoweeks; the second group received no exposure. Findingsdemonstrated thatnormally hearing adults can decode about 25% of He02speech asheard and that some iru&iduals are much more adept at the task thanare others. Second, it was found that a simpleexposure to He02 speech resultedin a neardoubling of decoding ability and familiarity with the talkers furtherenhanced this capability. INTRODUCTION Asmay be seen in Figure 1,it isnow well established thatthe combined effects ofthe 1!high ambient pressures, 2!helium/oxygen breathing gasmixtures and possibly! 3!high pressurenervous snyndrome, associated withsaturation diving, lead to severely distorted speech Bennett, 1967; Brauer, 1982; Brauer, etal, 1966; Fant and Sonnesson, 1964;Fant and Lindquist,1968; Fant, et al, 1971; Flower, 1969; Gelfand, etal, 1978; Hollien and Hicks, 1981; Hollienand Rothman, 1976; Hollien and Thompson, 1968; Hollien, et al,1973; Hollien and Hollien,1972; Holywell and Harvey, 1964; MacLean, 1966; Rothman, et al, 1980; Sergeant, 1963;Sergeant, 1968; White, 1955!. Indeed, even some ofthe specific speech distortions that haveresulted in theobserved communicative degradation have been identified Beil, 1962; Brubakerand Wurst, 1968; Hollien and Hicks, 1982; Hollien, et al, 1984; Hollien and Rothrnan, 1976;Hollien, et al, 1977;Rothman and Hollien, 1972; Sergeant, 1967!. Nese distorting effectsare further complicated by1! restricted system frequency response and 2! thehigh ambientnoise levels commonly found in habitats,chambers and the sea Brown, et al, 1977; Hollienand Rothman, 'l976!. In short,there appears to be aninverse relationship between 179 fearQ~nc&..1990 th 1 1of speech intelligiMity andhyperbaric dePth s e FIgure2!.This situation leapt, substanti'al communIca unicativedifficulties among aquanauts an4 with sup~~ surface. aPPoach p~y emPloyedto compensate for this debilitating involvedthe development and use of rnachine/computerspeech resto Wonymous,1988; Belcher, 1980; Belcher, 1982; Copel, 1966 G;11 1972. G 1973;Golden, 1966; Holhen and Rothman, 1972 Quick, 197O. R;ch~<s 198 t ods havebeen tried with varyingresults. included ha b P ~ r tec iques,time domain processing, digital coding oI 'o problems encount t e riM ~tems 1! tendnot to be speakerindependent and 2! o6'81 p spy . ForexamPle, the frequency domain systems op rateto 1 q ' whichhave been raised as the resultof the ~ of h h ! b reconstructthe spoken consonants. The di8iculty in this caseis that, while the thevowels ting to "carry"speech, it isthe consonants that contribute most heavilyto intelligibility. Moreotcr, thecited systems often "process" speech changes that arepsychologically motivated rather the dueto theeffects of thespeaking environment. Speaking fundamental frequencyis onemk factor;sometimes it is raised,sometimes not Hollien, et al, 1977!, yet most"unscratnbjen' ordinarilyanyway! tend to lowerit. Ia short,while someof the cited approacheswork fsirjl well-at least,under limited conditions-none of themappear to provide an acceptablesolutioI to the problem. Aninformal but rather widespread observation that can be made is thatsome &en andtopside personnel demonstrate what appear to be remarkabledecoding abilities 1+ speechdistorted byheliox gas mixtures andhigh ambient pressures. They are able todo % e"enin th«ace of1! noise Brown, etal, 1977!, 2!poor transmitting conditions iftolwi 4 3!th e fectsofthe sea ifsubmerged! andthe effects ofhearing loss ifpressurized Edrnn< Ho%enand Feinstein, 1976; Molvaer and Lehmann 1985!. Are these skills ~~ve~~ y madvertenttraining to our knowledge, noformal training pron of thLsNpe existed "-Arne'li ''~ty ' ' wtthtalker idiosyncrasies ~ ~ or some combination of thesefa ~,' Obvlousl «>nonis needed in order to determinewhich of theseelean< 5 controlling.As a matter fact't isessential todemonstrate thatthis phenomeno~ ex5"~ the Grstplace. Hollieii &.Thompson: Decoding Speech in HeliosEnv~nr PURPOSE The primary goal of this project was to discoverif simple exposureto distorted speech--specificallyutterances produced in the He02/Penvironment would enhancean auditor'sability to decodeit. Subjectiveobservation would suggestthat, while a predictionof this type is possible,there presentlyare no publisheddata on the issue.Secondary goals were to discoverif: 1! someauditors could be expectedto showgreater native ability re: the task than others,2! the greatestincreases in skill werethey to occur!would be demonstratedby thoseindividuals exhibiting the highestnative ability and 3! familiar voiceswould lead to an upgradeof decodingability. Of course,it wouldbe expected that some listeners would show superiortalent for thistask. However, it is notat all clearas to whetheror not theyalso would demonstratethe mostpositive response to "training".Nor wasit at all clear if individualswho wereexposed to particulartalkers and, hence, becaine familiar with them,also would show the greatestamount of improvement. METHOD Threeexperiments were carried out. Thefirst one was conducted in orderto testthe hypothesisthat exposureto He02/Pdistorted speech would result in improveddecoding accuracy.The second experiinent was a siinplereplication of thefirst and the third focused primarilyon thepostulate that auditors fainiliar with a talkerwould demonstrate enhanced ability to correctlyunderstand the speechuttered. Severaltypes of speechmaterials were utilized in theseexperiments. They included: 1!phonemically balanced lists seeCampbell, 1965! of 25words each, 2! theRainbow passage prose;99 words! and 3! exteinporaneousspeech. The word lists were used as stimuli when theexperimental or "training"!groups were exposed to theheliox speech at highpressures ie., He02/P!. Thetalkers utilized in aHthree experiinents were eight aquanauts who had participated in theU.S. Navy's "Sealab" program. They were selected from over 25 individuals who had beenemployed assubjects in a largenumber ofcominunication experiments carried out by theinvestigators. Tobe included inthese experimeiits, it wasjudged that 1! each talker should exhibitabout average speech intelligibility, 2!all talkers within a groupoffour exhibit similar scoresand 3! the two sets of talkers exhibit incan iiitelligibility levels that were approximately equal.Additional selection criteria specified that each talker be assigned toread only a single Campbelllist and that the mean score when doing so should reflect about a 257ocorrect intelligibilitylevel seeTable 1!. Mostwords were drawn f rom readings ina hyperbaric 181 Divingfor Science...1990 chamber!at a simulateddepth of 450fsw; however, in order to makeup equatedtests, some wordswere obtained from listsread at otherdepths, The other speechmaterials were drawn from diversspeaking at all depths.As maybe seenfrom Table 1, diverscores ranged from 24.4% to 30.7%with the meansfor eachgroup slightly abovethe desired25% and ap- proximatelyequal. These data were obtained by repeatedlyplaying the tapes to groupsof 10-15listeners who had normalhearing and someexperience with heliox speech.The final 100words produced by thefour aquanautscomprising Group A wereemployed to construct Test I and the 100words uttered by Group B constitutedTest II. The auditorsutilized in thisresearch were 64 young, healthy university students of both sexes.To be utilized asa subject,a volunteerhad to passa speechdiscrimination hearing test with a scoreof 92%correct or better. Twentyindividuals were randomly assigned to Experi- ment-1;20 to Experiment-2and the remaining24 to the third experiment. Thesame procedure was utilized for all threeexperiments. First, the subjects assigned to eachexperiment were randomly! divided into two equal groups;thus, there were two groupsof 10subjects each in Experiments-1and 2 andtwo groups of 12listeners in Experi- ment-3.While the groups were not controlled for sex,each consisted of roughlyhalf men and halfwomen. The equipment utilized to recordthe speech samples varied; it wasdictated by thatavailable for usein the particularexperiment from which that test materialwas drawn. However,it was pretty much state-of-the-science for the time-frame during which the experi- ment was conducted. The apparatusused in the listener procedureswas high quality laboratoryequipment. It consistedof a Ampex251 reel-to-reel tape recorder operated at 7.5ips,Marantz power amplifiers and two AR-1 speaker systems. All listeningsessions were conductedin a largesound treated room which had been specifically constructed for aural- perceptual experiments. Thesame procedures were utilized for all threeexperiments; Test I wasadministered to onegroup of the pair andTest II to the other. After the two-weekperiod had elapsed, subjectsin GroupA whopreviously had been administered Test I! respondedto TestII and thosein Group B whohad heardTest II! were administeredTest