Loudness of Complex Sounds As a Function of the Standard Stimulus and the Number of Components

Downloaded from orbit.dtu.dk on: Oct 02, 2021

Loudness of complex sounds as a function of the standard stimulus and the number of components

Florentine, Mary; Buus, Søren; Bonding, Per

Published in: Acoustical Society of America. Journal

Link to article, DOI: 10.1121/1.382062

Publication date: 1978

Document Version Publisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA): Florentine, M., Buus, S., & Bonding, P. (1978). Loudness of complex sounds as a function of the standard stimulus and the number of components. Acoustical Society of America. Journal, 64(4), 1036-1040. https://doi.org/10.1121/1.382062

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.  You may not further distribute the material or use it for any profit-making activity or commercial gain  You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Loudnessof Complexsounds as a functionof the standard stimulus and the number of components MaryFlorentine a) ENTDepartment, Gentofie University Hospital, 2900 Hellerup, Denmark qnd AcousticsLaboratory, Technical University of Denmark, 2800 Lyngby,Denmark S(/ren Buusb)

ß4coastitS Laboratory. Technical University of Denmark,2800 Lyngby,Denmark

Per Bonding ENT Department,Gentofie •niversity Hospital,2900 Hellerup, Denmarl( (Received15 lune 1977; revised10 April 1978)

The purposeof this studywas twofold:to determineif the measuredloudness level of a signaldepends on the standardstimulus used and to measureloudness as a functionof the numberof componentsin a wide- band signal.The stimuli were a pure tone, tone complexeswith frequencyseparations of 231 and 1592 Hz, and noi• bandswith widths of 220 and 1592 Hz. The centerfrequency was I kHz and the loudness level was approximately65 phons.Loudness matches between all combinationsof stimuli showedthat the measuredloudness of the soundsdid not dependon the standardstimulus used and the measuredloudness levelof a wide-bandsound increased as a functionof the numberof components.Individual observers were consistentin their loudnessestimations; the greatestsource of variabilitywas amongsubjects. Additional measurementsindicated that the rate at whichloudness incre. asexl beyond the critical band appeared to be greaterfor noisebands than for two-tonecomplexes.

PACS numbers: 43.66.Cb

INTRODUCTION tone complexes, and narrow-bandnoise. These stimuli have been compared in loudness to tone complexes or Zwicker and Feldtkeller (1955) showed that if the noise bands of various widths. The purpose of this p•- overall intensity of a noise band is held constant and per is twofold: to deterr•ine Lf the rate at which loud- bandwidth is increased, there is litfie or no change in ness increases beyond the critical band is dependent loudne.ssuntil the critical bandwidth is reached, after which loudness increases. The rate at which loudness upon the combination of stimuli used to measure loudness summation and to measure loudness as a function increases with bandwidth is most rapid at moderate in- of the number of components in a wide-band signal. tensities (Scha•f, 1959a; Zwicker and Feldtkeller, 1955; Zwicker• Flottorp, and Stevens, 1957). From I. EXPERMENT 1: LOUDNESS SUMMATION AS A. 1955 to the present, many experiments have examined FUNCTION OF STIMULUS COMBINATIONS loudness summation in order to measure the critical band. (For a review, see Scharf, 1970.) Althoughall A. Method these experiments are in reasonable agreement about I. Stimuli the width of the critical band at moderate intensities, they disagree about the rate at which loudness increases The loudnessof five soundswas investigated: a 1- with bandwidth or frequency separation beyond the crit- kHz pure tone, two-tone complexes with frequency sep- ical band. For example, some experiments reveal a arations of 231 and 1592 Hz, and bands of white noise slopeof 3-4 dB/ocf•of frequencyseparation beyond the 220 and 1592 Hz wide. All stimuli were centered geo- critical band for normal subjects (Niese, 1960; Port, metrically at 1 kHz. Limiting frequencies of the two- 1963; Zwickerand F•ldtkeller, 1955; Zwicker, Flot- tone complexes and the noise bands are given at the top torp, and Stevens, 1957), while other experiments re- of Fig. 1. The cutoff frequenci6sfor the noisebands veal slopes of 5-6 dB/oct at comparable intensities were the points at which the filter output was 3 ctB lower (Bonding, 1976; Florentine, 1977). The reason for this than the maximum output. All stimuli were 1 s in dura- difference in oblained slope beyond the critical band- tion with a rise-fall time of 35 ms. The fixed stimulus width is unclear. alternated with the adjustedstimulus; interstimulus in- terval was. 1 s. The componentsof the two-tone com- One difference amongexperiments is the type of plexes were set at equal intensity. Preliminary loud- stimulus used to measure loudness summation. The ness matches showedthat for each subject all compo- most frequently chosensland•rd stimuli are pure tones, nents were approximatelyequally loud whenequally in- tense.

a)Currentaddress: ResearchLaboratory of Electronics, 2. Apparatus Massachusetts Institute of Technology, Cambridge, MA 02139. Pure tones were generated by four oscillators (Hew- b)Currentaddress: Auditory Perception Laboratory, Fsychol- lett-Packard 200 CD), the -•-oct noise bandwas gen- og• Department, Northeastern University, Boston, MA 02115. erated by passing the output from a white-noise genera-

1036 J. AcoustSo• Am. 64(4}.Oct. 1978 0001-4966/78/6404-1036500.80 ¸ 1978Acoustical Society of America 1036 1037 Florentine,Buts, and Bonding: Loudne. of complexsounds 1037 tar(Br//el & Kjaer1405) through a r-octI filter (Brilel 4. Subiec= & Kjaer 1612), and a 1600-Hz noiseband was playedvia Ten subjects,six malesand four females, were test- a tape recorder (Telefunken M10A). The 1600-Hz noise ed. Half of the subjectshad previousexperience bandwas produced by taperecording (Telefunken M10 A) ing equal-loudnessjudgments and the otherhalf hadno the outputof a white-noise generator (Briiel & KJaer previousexperience except for three practicematches 1405), after filtering by a Krohn-Hite variable-band before' the onset of data collection. Ages ranged from filter (3750). Filters attenuatedabout 9.4 dB/oct beyond 20 to 36 years. Somesubjects were paid for their ser- the half-power points. vices. All subjectshad normal otoscopyand history Dependingupon the experimental condition, a pure and their thresholds were within 10 dB of ISO standard. tone, two-tonecomplexes, or noisebands were sent to one or bothchannels by meansof routing switches. Af- B. R•ult• snd di•cu•ion ter amplification, the continuoussignals were gated, Results for all 25 matches by all ten subjects are attenuated,and then mixed. Bothchannels were identi- summarizedin Fig. 1. The level differencesneeded to cal except for a sonspotentiometer with a 60-dB range obtainequal loudness between the standard stimulus and .in the variable channel. The same Lestton timer which thecomparison stimulus are shownfor all combinations triggered the gates also tribered a shaper positioned of the five stimuli. The horizontal lines indicate the after the mixer. This arrangement assured that both meanvalues, the whitebars representthe interquartile signals had identical shape. Next, the signals were at- rangesand the black bars representthe total ranges. tenuated and led to a TDH-39 earphone mounted in a MX-41/AR ear cushion. The stimuli were measured An analysis of the means and the covarlance was per- andmonitored by anelectronic counter (Hewlett-Pack- formed on linear combinations of the data. The results ard 5223L), an oscilloscope(Tektronix 5103N), and a of theX- Y matchand the Y- X matchwere summed voltmeter (Br/iel & Kjaer 2603) which measured the rms for each individual observer. The mean of these linear voltages. combinationswas not significantly(p < 0.05) different from zero for any combinationof two signals. This showed that the results for each observer were sym- 3. Procedur• metrical. In other words, it did not matter which sig- In accordancewith a matrix design, eachof the five nal was the standard stimulus and which signal was the signalswas matched in loudnessto itself andto eachof comparison/•timulus. the other•ignais. The 25 matcheswere run in random Furthermare, a two-wayanalysis of variancere- order, with a different randomorder for eachsubject. vealed that observers show transitivity among all loud- Subjectswere instructedto pay attentiononly to the total ness matches. For each observer the sum of the level loudnessof the sigeals. First, the subject matchedthe. differences neededto obtain equal-loudn•.se among any otherfour si/nals in loudnessto the i-oct noise'band three signals approximateszero. Transitivity was con- set to 65 dB SPL. Then each signal was matched in firreed by the fact that the variability amongsubjects, loudness to itself and to each of the other signals. For obtainedby poolingdata for an X- Y matchfor all sub- these matches each standard stimulus was set to the in- jects, wassignificantly (fi < 0.05) greater thanthe vari- tensity obtainedin the first four matches in order to ability within subjects, obtainedby poolingthe results keep the loudnessconstant. Matches were made mon- of the matches over different standard stimuli from one aurally by the method of adjustment. The subjects were subject. instructed to bracket the standard, i.e., to set the ad- Jusledstimulus alternately louder and softer than the Far all standardstimuli measured, loudnessfor the fixed stimulus, reducingthe differenceuntil they per- wide-bandnoise was greater than for the wide-bandtwo- ceived equal loudness. After each judgmentthe experi- tonecomplex. This differencewas also foundwhen fil- menter cha•ed the attenuationin the variable channel ters with slopesof approximately200 dB/oct were used. in order to prevent position cuss on the subJect'sinten- On the average, subjectsrequired the wide-bandtwo- sity control knob. Four judgmentswere madefor each tone complexto be 10-11 dB more intensethan the wide- stimulus pair. The comparisonstimulus was adjusted bandnoise to obtain-equalloudness. This difference twice and the standard stimulus was adjusted twice. H wasstatistica.l.ly .independent of whichstandard stimulus thedifference between two judgments of identical stimu- was used. lus configuration was greater than 3 dB, the j.udgment The findingthat the measuredloudness of a wide-band was repeated until two Judgmentswith a difference less noise was greater than the measuredloudness of a two- than 3 dB were obtained. For approximately 95% of the tonecomplex appears t0be in conflictwith an earlier matches no extra judgmentswere required. In approxi- studyby Scharf(1959b). He showedthat the loudness of mately 5% of the matches one extra Judgmentwas re- tone complexeswith widthsof 1600 and 3400 Hz at a qulred; only a few matchesrequired two or more extra center frequency of 1500 Hz did not dependon the num- judgments. Both the comparisonstimulus and the stan- ber of components. The reason for this difference is dard stimutus were varied in order to cancel t• tenden- unclear. cy to set the adjustedstimulus to a hi,her intensity rela- tive to the fixed stimulus. (See Scharf, 1961.) Blotime The presence of symmetry and transitivity, in the limit was imposed, and the subject usually listened to a present data, indicatevery COnsistentand orderly re- pair approximately20 times before reporting a match. sults that are not an artifact of the stimuli and method 1038 Florentine, Buus, and Bonding: Loudnessof complex sounds 1038

STANDARDS

891 1122 891 1122 1000 /,68 2060 468 2060 Hz

•F=220 Hz AF= 231 Hz /•F =OHz AF=1592Hz ,,'XF=1592 Hz A B C D E 30

•' 20

:z o u• 10

A S C O E A B C 0 E

z ABCDE

-2O

A B C 0 E RANGE

¾ ¾ C: ¾ 0: •F: 220 Hz /•F :231 Hz •F=OHz AF = 1592 AF=1592 Hz FIG. I. Leveldifferences needed to obtainequal loudness between the standar d stimulus and each of thecomparison stimuli for the five standard stimuli. Each bar represents a total of 40 judgments by ten subjects. The horizontal lines. white bars. and black bars represent the medians, interquartile ranges. and total ranges, respectively. used in this experiment. Furthermore, we have repli- II. EXPERIMENT2: LOUDNESSSUMMATION AS A cated some of the results of the present experiment with FUNCTION OF THE NUMBER OF COMPONENTS different subjects in three different laboratories using Experiment 1 suggested that the measured loudness two different psychophysical procedures. of a wide-band sound may increase as a function of the number of components. The purpose of experiment 2 C. Variability was to measure the loudness of white noise and of mul- Correlations between the loudness estimations of the titone complexes composedof two, three, and four signals for each subject revealed that subjects were - tones. consistent in their judgments of all the stimuli. In other words, if a subject needed a large level difference A. Method between two signals to obtain equal loudness when one Five observers were used, four of whom participated standard stimulus was used, he also needed a large dif- in experiment 1. The components of the test stimuli ference when another standard stimulus was used. were evenly spaced in frequency around a geometric When a signal was matched in loudness to itself, varia- mean of 1 kHz and set to equal intensity. Each com- bility was approximately the same for all five signals. parison stimulus was matchedin loudnessto •a220-Hz More variability was obtained when two different signals noise band by the method of adjustment as described in were matched. When the wide-band noise was matched experiment 1. in loudnessto any of the four other soundssignificantly (p < 0, 05) greater variance was obtained than when B. Results and discussion matching any combination of the other four signals. The fact that variability amongsubjects was significant- The individual data from the five observers are shown ly greater (p < 0.05) than that within subjects shows that in Fig. 2. For all subjects the measured loudness in- the individual subjects needed different level differences creases as a function of the number of components. to obtain equal loudness. There are two possible reasons why Scharf (1959b) found

J.Acou•t. Soc. Am., Vol. 64, No. 4, October1978 1039 Florentine,Buu•, end 8or•i•: Lo.d".• of complexmuncl• 1039

monic. The methnd of maximum likelihood, an ada•- LOUDNESS AS A FUNCTION OF THE NUMBER OF COMPONENTS tire forced-choice procedure (Lyre•aard and Pealersen, CENTER FREQUENCY,, 1000 Hz 25 1971), wu used to set the sig•is equal in loudness. STANDARO:NARROW BANO NOISE AT 65 d8 SPL .e ß St Each match was based on approximately 25 responses. o S2 ß (For further details, see Florentine, 1977.) A S3 B. Rmultsand discuuio• Individualdata from twoobservers axe shownin Fig. 3. Each point represents the median of three equalloudness ma•ches. Subjects reported that they per- ceived one sound image •nd that they tried to base their jud•nents on the tot• loudness. Results from both subjects showthat loudnesssumn•tes more rapidly for noise bands than for two-tone complexes even at fre- 2-TONE 3'•'ONE &-TONE NOISE WHITE h----COMPLEXES----I BAND NOISE quencyseparations clo•e to the critical bandwidth. ! -'-F=1600 Hz TEST STIMULI iV. SUMMARY FIG. 2. Level difference needed to obtain equal loudness as a function of the number of componentsfor five subjects. Each (1) Loudness summation did not chan• with the dif- point represents an equal loudness ma•ch by the method of ad- ferent standard stimuli used in experiment 1. Justment. (2) The rate at which loudness increases beyond the critical band appears to be greater for noise bands than no loudness summation as a function of •-e number of for two-tone complexes. components and we did. The first concerns the treat- (3) While the overall bandwidthand intensity of a ment of the data. Loudnessmatches by different ob- wide-band stimulus is held constant, loudness increases servers have generally been pooled. Furthermore, as components axe added. sometimes the same subjects have not served under all experimental conditions. Since the variability amon• (4) Individual subjects were. consistent in their loud- subjects is laxger than the variability within subjects, ness estimations. Differences among subjects was the the loudness difference among stimuli could be ob- •Teatest source of vaxiabfiity. scured. A second reason is that Scharf did not test his (5) The greatest variability was obtained when the subjects at 65 dB SPL, where the maximum loudness 1592-Hz noise band was matched in loudnens to another summation occurs. Since the total effect of loudness signal,primarily owing to intersubjectvaxi/bility. summationwas small •md v•riability was large, the difference in loudness summation could not be seen.

It is interesting to note a similarity here between the measured loudness and the acoustic reflex: Threshold LOUDNESSSUMMATION AS A FUNCTION of the acoustic reflex also depends on the number of 20 OFBANDWIDTH componentsin a wide-band sound (Popelks, K•rlovtch, 15 ß NOISEBANDS and Wiley, 19'/4). • o 2-TONECOMPLEXES •'•

III. EXPERIMENT 3: LOUDNESS SUMMATION AS A FUNCTION OF BANDWIDTH FOR TWO-TONE COMPLEXES AND NOISE BANDS •o (,2 Experiments 1 and 2 suggestedthat the loudnessof a 1592-Hz wide-band si•mal increases as a function of the ' I0 ,', number of components. The purpose of exl•riment 3 was to examine if the rate at which loudness increases beyond the critical band is sreater for noise bands th•n for the two-tone complexes. o

A. Method 50 •00 200 500 lOO0 2000 Loudness summation as a function of bandwidth was FREQUENCY SEPARATION (/,F) IN Hz measured for both noise bands and two-tone complexes, FIG. 3. Level dLffe•enee m•

J. • S0C.Am., VoL 64, No. 4, Octob• 1978 1040 Florentine,Buus, and Bonding:Loudness of complexsounds 1040

ACKNOWLEDGMENTS Level of Impulsive Noise," Rep. No. 6, Acoust. Lab., Tech. Univ. Denmark. We are grateful to Professor Gerhard Salomon and Ntese, H. (].960). "SubjectiveMessung der Lastst'•rkevon lecturer O. Juhl Pealersenfor obtainingfunding and Bandpassrauschen, Hochfrequenztech.Elektroakust. 68, providing excellent working conditions, to lecturer 202-217. Torben Poulsen and civil engineer Claus Elberling for Popelka, G. R., Karlovich, R. S., andWiley, T. L. (1974). recording the noise band tapes, to SusanneM•ller for "Acoustic Reflex and Critical Bandwidth," J. Acoust. Soc. statistical assistance, and especially to Professor Ber- Am. 55, 883-885:. tram Scharf and Professor Eberhard Zwicker for read- Port, E. (].965). "Uber die Lautst•rke einzelner kurzer Sohall- impulse," Acusttea 15, 212-223. ing critically an earlier version of the manuscript. Scharf, B. (].959a). "Critical Bands and the Loudness of Com- This study was supportedby grants from the Research plex SoundsNear Threshold," J. Acoust. Soc. Am. 81, 365- Medical Council and the Tuborg Foundation. Research 380. was carried out while the first author was on leave Scharf, B. (].959b). "Loudnessof Complex Soundsas a Func- from the Auditory Perception Laboratory, Northeastern tion of the Number of Components," J. Acoust. SOc. Am. 31, University, Boston, MA 02115. 783-785. Scharf, B. (196].). "LoudnessSumm•.tion under Masking," J. Acoust. Soc. Am. aS, 503-511. Scharf, B. (1970). "Critical Bands," inFoundat/onsof Modern Bonding, P. (1976). "Critical Bandwidth in Persons with Auditory Theory, edited by J. V. Tobias (Academic, New Hearing Loss Due to Salicylate," Proc. 12th Int. Congr. York), ¾ol. 1, pp. 157-202. Audiol., Florence, Italy, 142-143. Zwicker, E., andFeldtkeller, R. (].955). ' '•er die Lantst'•rke Florentine, M. (].977). "The Effect of PsychophysicalProced- yon gleichfSrmigenGer'•uschen," Acustica 5, 303-316. ure on Measurements of the Critical Band by Loudness Sum- Zwicker, E., Flottrop, G.• and Stevens, S.S. (].9{17). "Criti- mation," Rep. No. 20, Acoust. Lab., Tech. Univ. Denmark. cal Bandwidth in Loudness Summation," J. Acoust. SOc. Am. Lyregaard, P. E., and Pedersen, O. J. (].97].). "Loudness 29, 548-557.

J. Acou•t. Soc. Am., Vol. 64, No. 4, October 1978