Dept. for Speech, Music and Hearing

Quarterly Progress andStatus Report

The acoustic properties ofvoice timbre types and the

importance of theseproperties in the

determination of voiceclassification in male singers

Cleveland, T.

journal: STL-QPSRvolume: 17number: 1year: 1976pages: 017-029

http://www.speech.kth.se/qpsr

STL-QPSR 1/1976 17.

C. THE ACOUSTIC PROPERTIES OF VOICE TIMBRE TYPES AND THE IMPORTANCE OF THESE PROPERTIES IN THEjtDETERMINATION OF VOICE CLASSIFICATION IN MALE SINGERS

Abstract

An investigation to determine the significance of formant frequencies, pitch, and source spectrum slope on voice classification is reported. Eight professional s ingers sang five common vowels on four common pitches, and, i n a forced-choice test , vocal pedagogues classified the vocalizations as tenor, baritone, or bass. Source spectrum slopes of 12 dB per octave approximated the vocalizations of the three voice c lasses , while formant frequencies and pitch accounted for most of the difference in jury evaluation resul ts . To evaluate m o r e accurately the significance of formant frequencies, pitch, and source slope on voice classification, vowel synthesis was employed. Representative voice c lass formant frequency combinations, four different pitches, and source spectrum slope were systematically al tered while, vibrato, and vowels were stabilized. The original jury classified the synthesis in a forced- choice test , and the dependence of the jury evaluation upon formant f r e - quencies and pitch was evidenced in analysis of variance tests . Addi- tional experiments based on the above findings suggested the following resu l t s and conclusions:

I ) Measurements of spoken vowel vocalizations revealed higher formant frequencies in tenor t imbre type and lower formant frequencies in bass t imbre type.

2) The "average formant frequency" seems to be a better description of t imbre type than separate , individual formant frequencies.

3) Long-time-average- spectra seem to afford a good est imate of voice classification.

4) Formant frequency percentage differences between basses and tenors were s imilar to those found between ma les and females suggesting that the mouth and pharynx lengths known to differ entiate females and males , probably a l so differentiate tenors and basses .

.jC This paper i s a comprehensive overview of the author 's doctoral dissertation presented to the University of Southern California. The r e sea rch and writing was accomplished during 1973- 197 5 while the author was a guest r e sea rche r a t KTH, Dept. of Speech Com- munication under the sponsorship of the Fulbright-Hays Foundation.

* Author' s present address: Laryngeal Dynamics and Vocal Acoustics Laboratory, University of Southern California, School of Medicine, R e search Division, Department of Otolaryngology, 903 Hoffman, 2025 Zonal Ave., L o s Angel e s , California 90033, USA.

STL-QPSR 1/1976 18.

Introduction

Voice timbre i s often defined a s that particular attribute of a given

voice which distinguishes that voice from another when the vowel and the

pitch a r e the same. Individual voice timbres may be said to exist along

a continuum of a myriad of variour timbres. Vocal pedagogues have em-

pirically divided the continuum of various timbres into a t least three

main timbre divisions: bass, baritone and tenor. Acoustically, in-

dividual timbre s would correspond to characteristic acoustic signals of

the laryngeal source and the vocal t ract resonances, while voice timbre

types, i. e. , bass, bariton and tenor, would be groups of individual

timbres which possess similar laryngeal source and vocal t ract reso-

nance characteristics.

Because no cataloging of acoustic characteristics of timbre types

exists, the realization of such a catalog might assist in the under standing

of certain required acoustic properties of timbre types and, more im-

portantly, the ascertainment of the importance of timbre type in the

determination of voice classification might ass is t in the development of

a theory of voice classification for both speaking and singing voices.

The present work i s intended to contribute to such a theory.

Scope of this study

The present investigation encompasses a ser ies of experiments in-

volving voice timbre and voice classification in male singer s. Fir st,

the study seeks to ascertain the acoustic properties which a r e influential

in the perceptually recognized timbre type s of bass, baritone and tenor.

Second, the study probes the importance of timbre type in the determina-

tion of voice classification. Third, the study presents rough calculations

of "typical" timbre type vocal tract lengths computer from formant fre-

quency measurements.

Experiment I - Real Sounds

Method ---- In order to determine the spectral characteristics of timbre types,

eight professional Swedish male singers sang the vowels [ i ] , [ e ] , [ a 1, [o 1, and [ u ] in a prearranged scheme on the pitches C

3' F3' Ag and Eq. The source spectra and formant frequencies of these vocal-

izations were determined by analysis by synthesis on a terminal analogue

of the vocal tract. The formant frequencies of lower pitches were also

determined by sonogram measurement.

STL-QPSR 1/1976 19.

A tape of commonly spliced (common onset/decay), randomly ordered,

vowel vocalizations was prepared for the perceptual evaluation of eighteen

vocal pedagogues. The individual vocalizations were presented three con-

secutive t imes with s ix seconds of silence following the l a s t presentation

affording the ju ro r s an opportunity to classify the sounds a s bass , bari-

tone o r tenor.

The listening tes t which was played to the jury on a tape r eco rde r

(Revox 477) was heard by the jury via headsets (Sennheiser HD 44). Be-

I

cause of the la rge number of sample sounds ( 5 vowels x 4 pitches x 8 sub-

jects) the tes t was divided into two sessions of approximately 25 minutes

each. An interval "break" of 30 minutes separated the two sessions.

The jury was instructed to not discuss the tes t during the "break".

Conversion of jury response to jury number

A number representative of the jury evaluation (JN) of each sound

stimulus was computed f rom the formula 2 + lBr - Z B = JN, where T,

Br, and B a r e the voice t imbre types Tenor, Bariton and Bass, respec-

tively. -k

Results ----- a. Formant frequencies

Correlation of the average formant frequency to the mean jury num-

ber of each subject may be seen in the sca t te rgram of Fig. I-C-I. The

average formant frequency i s an average of the four lowest formant f r e -

quencies of the combined vowels [ i ] , [ e 1, [ a ] , [o 1, [u ] of each

singer - subject and the mean jury number i s the mean of the jury numbers

of a l l the vocalizations by an individual singer - subject.

A correlation coefficient of 0. 90 was calculated showing that a change

in the mean jury number corresponded closely to a change in the average

formant frequencies of the singer-subject. The statist ical significance

of the correlation coefficient was tested by transforming i t to a "tl1 statis-

tic. The correlation coefficient of 0.90 corresponds to a "t" statist ic of

* The formula IT + 0 - iB was felt to reflect the l inear relationship which probably exis?: between voice t imbre types in the remaining experiments of this investigation so i t was employed. However, r e su l t s f rom this new formula did l i t t le to change the resu l t s f rom the previous formula except lower al l jury numbers by 7 jury num- b e r s with a standard deviation of + 1 jury number.

MEAN JURY NUMBER AS A FUNCTION OF AVERAGE

FORMANT FREQUENCY Ci ,e,a,o, ul

- % 0

- . . . . . . . . . .

.

9

. . . . . . . . . . . . . x HH

9 X JA BE

I I I

1.6 1.7 1.8 AVERAGE FORMANT FREQUENCY (kHz)

Fig. I-C- I.

6.94, which, with six degrees of freedom, i s significant at the .001

level of confidence. From these results i t seems reasonable to assume

that one of the acoustic properties on which the jury evaluation was de-

pendent was formant frequencies.

It i s of significant interest to note the grouping of the singer-subjects

as shown in Fig, I-C- I . It appears that the jury placed the singer- sub-

jects into three different groups. The groups may be readily distinguished

in the figure. The lowest group i s comprised of HH and BE; the second

group i s made up of JA, SS and UB; and the third group i s comprised of

SEA, LS and CS. Furthermore, because the groups a r e so explicitly

delineated, i t may be assumed that the jury used a criterion which i s re -

lated to the average formant frequency. * Average formant frequency I

boundaries between the groups might be appropriately drawn at I. 69 kHz

(bass and baritone) and a t I. 79 kHz (baritone and tenor).

b. Pitch

In order to determine if the different pitches elicited a different res -

ponse from the jury, the mean jury number was correlated to the various

pitches. In this case, computation of the mean jury number was derived

from the average jury numbers of all vowels from a single pitch by a

given singer- subject. For purposes of correlation, pitches were defined

by their fundamental frequency. The results of this correlation may be

seen in Fig. I-C-2. The scattergram shows that the jury votes changed

with a concurrent change in pitch. A correlation coefficient of .81 was

computed for these data, and i t seems reasonable to infer that pitch was

also an influential acoustical property in the jury 's determination of

voice timbre type.

c. Source

As the timbre type may be dependent on the source spectrum, the I I

source spectra of the singer-subjects in this investigation were studied.

Resulting measurements from the study determined little consistent de-

viation from a source falling a t - 12 dB per octave for the various singers

* An extensive examination of several representative values of formant frequencies suggested that the "average formant frequency" a s utilized in this experiment was a representative of the formant frequencies a s any number that was explored. A similar number has been used in other investigations, e. g. Coleman, 1973.

JURY NUMBER AS A FUNCTION OF PITCH

[i.e,a,o.u I

v CS A LS 0 UB

SS 0 SEA x HH

JA

PITCH (SEMITONES)

Fig. I-C-2.

SOURCE SPECTRUM SUMMARY

measured - synthesis (Tenor) - a sinthesis (Bass!-x Harmonic number.

a. d B ! . . . . . 5 .... . . .a . lo . '? ....

I I . -

Frequency -

Harmonic number . ...... .. 5 10 15

L. ..--.a ., 1 1&0

I Frequency

I =- Harmonic number

1 dB .A,-- --. 10 15 .............. - 5. . . . . . . . . a ....-.-,. 1 I . . . . .

..... - - . . . . . . . . . . . . . . . . . . . . . . . * . . . . . ... --.--- ----..-- I- --- '------- ,.----I

329 1645 . 3290 4 9 s Hz F reouencv

Fig. I -C-3 .

STL-QPSR 1/1976 22.

viewed somewhat tentatively, however, a s there a r e many uncontrolled

variables in rea l sounds. More definite conclusions may be drawn i f the

acoustic proper t ies which seem important in the generation of t imbre

types a r e synthesized and aurally evaluated by the same group of vocal

pedagogues.

Experiment 2 - Synthesis -

Method ---- A ser ies of vowel sounds were synthesized which reflected the a-

coustic properties found to be influential in timbre type determination of 1

rea l sounds. The synthesis was accomplished utilizing the source-filter

network shown in Fig. I-C-4. Formant frequencies and pitch were sys-

tematically varied while the source spectrum and vibrato were kept con-

stant. A synthesis evaluation tape of the synthesized sounds, which was

similar to the evaluation tape of the rea l singer - subject vocalizations,

was made and exposed to the original jury of vocal pedagogues, who

classified the sounds a s bass, baritone or tenor. The formant frequen-

cies were derived from coordinates of a best fit line through formant

frequencies of individual formants and equally spaced points along a

continuum of jury numbers. The best fit line was drawn through a scat-

tergram of formant frequencies of a given formant from all singer sub-

jects on all pitches and the equally spaced jury numbers along a conti-

nuum were chosen by the investigator a s representative of the various

timbre types. A scat tergram example from which data points were de-

rived a r e given in Fig. I-C-5. Pitches were the same a s those used by

the singer- subjects and the vibrato was f 2 OJa of the fundamental. This

vibrato generation i s consistent with human vibrato sounds ( ~ o l f e et al,

1965).

I Results -----

The resul ts of the jury evaluation were statistically measured by

univariant arlalysi s of variance (Finn, 1974) on the representative vowels

[i j [ cr, ] [ u 1. The resul ts of this test a r e shown in Table I-C-I. The

statistical F values (F) for the variable formant frequencies, here r e - f

ferred to a s formant type, and pitch a r e both significant a t the . 0 0 1 level 1 I of confidence. 1

I

As can be seen in Table I- C-I an additional test of orthogonal poly-

nomial contrast was also accomplished. Such a test determines how the

VOWEL SYNTHESIS

I (vibrato) I

SOURCE - FILTER DIAGRAM

low f reqwncy

SOURCE

sine wave

( monitor 1 Spectrograph rl i I

Fig. I -C-4 .

. * I

Low Pass i 5,F2JF3 f 4. F 5. . m, Filter a I HP)

pulse I a I I Formant

train I I Synthesizer 1

Sine Wave n- Generator .

I I I

Pulse Generator

RELATIONSHIP OF JURY NUMBER To

FORMANT FREQUENCY ta1

JURY NUMBER

Fig. I-C-5.

TABLE I - C-I. Univariant analysis of variance resul ts

Source d f ----- Constant I - - Formant type 2 1621.75 36 .56 .0001

(lin. ) ( I ) (3243. 37) (73. i I) . 0 0 0 1

(quad* ( 1) (. 125) (. 003) .9582

Pitch 3 3881.29 87.49 . 0001

Interaction 5 175.45 3. 96 .0069

Within 24 44. 36

- Totals 36

Univariant analysis of variance test resul ts showing

highly significant dependence of jury note on the synthesis

treatments, formant type, pitch and the interaction of the

two. Furthermore, orthogonal polynomial contrasts show

formant type treatments to be linearly related with a sta-

t istical significance beyond the . 0 1 level of confidence.

STL-QPSR 1/1976 23.

means a r e related. For example, i t asks the question, Do the means of

the different treatment groups, a s in this case, formant type, have a linear,

quadratic or cubic relationship? As can be seen in the table the statistical

F value i s significant at the . 0 1 level of confidence for the linear (lin) con-

t ras t of means. This result indicates that the means of the formant type

treatments-bass, baritone and tenor a r e different, and their differences

fall on a straight line. Fig. I- C- 6 illustrates the linear relationship be-

tween mean jury number and formant type for the vowels [ i, a, u ].

Table I-C-I also shows the interaction of the two variables, formant

type and pitch, a s tested by the univariant analysis of variance. The in-

teraction is significant a t the .006 level of confidence.

Discussion ------ It i s interesting to compare the results obtained from analysis of the

real sounds and the synthesized sounds. Where in the real sounds the

formant frequencies seemed to be a more important acoustic property

than pitch, the reverse seems to be the case with the synthesized sounds.

Possible explanations for this difference may be a s follows: I) For anal-

ysis, the jury mmbers from the synthesized results were "normalized"

with respect t6 pitch and formant type, whereas "normalized" results in

the real sounds were not feasible. Thus, llnormalized" results were not

compared to "normalized" results. In addition, 2) The crude method

utilized to derive the formant frequencies for the synthesis may be res -

ponsible for some of the difference.

Conclusions ------- Regardless of the difference between the real and synthesized statis-

tical results, the results a r e compatible and suggest that it i s reasonable

to conclude that timbre type i s dependent on formant frequencies and pitch.

However, we may regard the influence of pitch on timbre type determina-

tion a s a property more of theoretical interest rather than practical im-

port, and, a s such, from a practical and pedagogical point of view, i t I

may be concluded that timbre type i s most practically influenced by for-

mant frequencies.

Experiment 3 - Practical import of timbre type

Introduction and method ------------- Since voice timbre type seems quantifiable through the utilization of

formant frequencies or i t s counterpart the average formant frequency,

a s shown in Fig. I- C- I, the second purpose of this investigation - deter-

mining the practical import of timbre type in voice classification was

explored. To do so, i t was necessary to ascertain the importance of

other properties important in voice class determination. Range i s such

a property. The center range was determined by finding the range of a

given singer-subject on a piano with 81 keys and assigning the most

central pitch in that range the number of that key a s they were counted

from bass to treble. This logarithmic number was preferred over the

linear number which would have prevailed i f the pitch frequency had

been employed. Fig. I-C-7 shows the correlation of the center range

with autoclas sification (the professional singing classification of the I I

singer s in this investigation).

Results ---- I

The correlation coefficient computed for these data was .97 suggest-

ing an almost perfect correlation between center range and autoclassi- 1 fication. From this result i t seems reasonable to assume that, practically

speaking, the center range might be regarded a s indicative of the auto-

classification of a professional singer. Consequently, a high correlation

between center range and average formant frequency would suggest that

an estimate of the autoclassification might also be derived from the

average formant frequency alone.

Fig. I-C-8 is a scattergram of center range and average formant

frequency (the number utilized in this experiment to quantify timbre).

The plot supports the conclusion - the higher the center pitch of the

range, the higher the average formant frequency - though there i s one

case in which this does not hold - singer subject HH does not conform

to the trend in the remaining data. In spite of this result, however, the

finding seems to suggest that, a s a rule, the formant frequencies and

center range tend to "matchu, and since center range seems to be a good

indicator of professional voice classification, formant frequencies or the

average formant frequency of a singer-subject might, therefore, be a

good indicator of professional singing classification.

Further assistance in classifying voices might be gained i f i t could

be determined that formant frequencies in speech correlated to auto-

classification. Results from an extensive study of speech sounds of the

professional singers in this experiment reflect that formant frequencies

of speech sounds correlate highly to au toclassification. Measured

TEN

U tZ

BAR

AUTOCLASSIF1 CATION AS A FUNCTION

OF CENTER RANGE

I I I I

30 35 40

CENTER RANGE (pitch number)

Fig. I -C-7 .

CENTER RANGE AS A FUNCTION

OF AVERAGE FORMANT FREQUENCY

UB SS SEA

1.6 1.7 1.8

AVERAGE FORMANT FREQUENCY kHz)

Fig. I-C-8.

formant frequencies from several vowels were correlated to the auto-

classification of the singer- subject s and the resulting correlation coef-

ficients a r e given in Fig. I-C-9. As seen in the figure, the high corre-

lation coefficients seem to indicate that measurements of formant f re-

quencies of speech vocalizations may be helpful in voice class deter-

mination.

Discus sion c-----

The results of these experiments seem to be valuable for singers

and pedagogues alike for the findings may offer explanations to the peda-

gogical observations heretofore unexplained. Pedagogues have often

made the observation with reference to untrained singing voices, "If the

quality i s there, the range will come". It seems possible to suggest an

explanation from the evidence in this study a s to what a teacher means

when he makes such a statement.

Range i s usually a product of vocal maturity and i s normally fully

extended only in trained singers. On the other hand, this experiment

has shown that voice timbre i s related to formant frequencies which I

I I

depend on vocal t ract size and dimensions (Fant, 1960). If timbre is a

result of vocal tract size, i t may be suggested that timbre would exist

at an earlier age than range in the beginning singer, a s i t would be ex- i

pected that a singer would reach a more physically mature state than

that stage a t which his range would be more fully developed. Therefore,

timbre may be a better early estimator of voice class in the beginning

singer than range. The pedagogue who has made this observation, em-

pirically, has possibly recognized through year s of observant voice

teaching that timbre in the untrained voice i s a good indicator of the

future voice class of the singer because, in professional singers, a s I

shown in Fig. I-C-8, the timbre and the range tend to llmatch". The

same explanation may even apply for the more mature singer who pos-

sesses a wide, ambiguous range, encompassing more than that required

for single voice class. In determining the voice class of this singer,

the pedagogue may be seeking the best "match" between timbre and range

utilizing timbre a s an empirical cue to assign the appropriate voice class.

For the pedagogue who has witnessed a singer who possesses baritone

range, but tenor quality, a similar explanation may hold. Such an im- pression may be evoked by the empirical observation that a given voice

STL-QPSR 1/1976 26.

i s a "mismatch" voice. As this experiment shows that, generally, i f

the formant frequencies a r e high, the range may also be high, the peda-

gogue may be expressing empirically the same point that timbre and

range should "match".

Conclusion ------ It seems in order to conclude that in the absence of an extended range

of pitches, and in the presence of an ambiguous, extensive range of pitch-

es, timbre may be a good estimate of the future voice class of a given

singer.

Experiment 4 - Long- - . - Time-Average-Spectra ..- . -

Introduction and method ------------- Since the finding of this experiment seems to have much practical

value, a more expedient method than that of sonogram measurement and

spectrum matching should be realized. Such a method may exist in long-

time-average-spectra. LTAS i s a single spectrum of the averaged a-

coustical components of short or extended signals a s a function of time.

In speech LTAS peaks a r e located around the average formant fre-

quencies of the three lowest formants ( ~ i i l l e r , 1973). In singing, three

peaks generally appear, but the third peak, the singing formant peak, i s

more pronounced ( ~ a n s s o n and Sundber g, 1972). I t would be anticipated

that singers singing the same task on the same pitch would reflect char-

acteristic "timbre" spectra.

Results ----- Such i s the case in Fig. I-C- 10 which reflects a comparison of LTAS

of two professional singers singing the same vowel scheme on the same

fundamental frequency. Singer- subject BE' s autoclas sification i s bass,

and he was classified by the vocal pedagogues a s possessing bass timbre.

Singer - subject LS has the professional singing classification of a tenor and

was classified by the jury a s a singer with tenor timbre. The figure

clearly delineates the timbre differences.

Conclusion ------ The results from this LTAS, and a systematic study of different LTAS

on other tasks allows the conclusion that LTAS holds much potential in

the ascertainment of timbre type and voice classification.

.- SO DB BE and LS. vowel scheme, A3

-BE - LS

Superimposed LTAS o f bass (BE) and tenor (LS) showing consequential spectral d i f fe rences resul t i n g from the s inging o f the same vowel scheme on the comnon p i t c h Ag.

Fig. I-C-10.

STL-QPSR 1/1976 27.

W e r i m e n t 5 - Morphology

Introduction _------ Since this study has shown characteristic formant frequency patterns

for bass, baritone and tenor timbre, morphology of the vocal t ract must

be related to voice t imbre type and voice classification. Studies of various

investigators have shown that, in most vowels, the formant frequencies

of women and children a r e considerably higher than those of males (Fant,

1960; Peterson and Barney, 1952) and that these differences a r e due to

the relatively greater pharyngeal length and more highly developed la-

I ryngeal cavities of adult males over that of females, and that same phy-

sical maturity of females over children. However, within these groups I

no systematical subdivisions seem to have been established.

Results ---- Characteristic formant frequencies typical of the voice timbre type

(as determined by the jury of vocal ~edagogues) and the autoclassification

(professional singing classification) of the singers in this study a r e given

in Table I-C-IIa and b, respectively. The characteristic frequencies a r e

presented by vowel and represent an average of measured formant fre-

quencies of all singer- subjects who possess a like timbre type a s a

result of jury classification or a similar voice class a s a consequence

of autoclassification.

Scattergrams of the correlation of the formant frequency difference

(measured in per cent) between males and females and between auto-

classified basses and tenors, and jury classified bass timbre and tenor

timbre i s shown in Fig. I-C- 1 I. The resul ts were obtained using the 'it formula Fn% A = (F - 1) loo%, where F equals the formant frequency in

lb question, "in i s the formant number, and "t'l and "b" equals tenor and

bass, respectively. Comparison of the percentage differences between

jury classification (dashed line) and autoclas sification ( solid line) with I female/male (line-dot) of Fant' s study (Fant, 1966) shows a remarkable

correspondence in most cases. Especially close relationships a r e seen

in FI of [ a ] and [ o ] and F2 of [ i , e , a , u]. A s can be seen in the

Figure, the largest percentage differences between females and males

a r e also the largest percentage differences between tenors and basses.

Since i t i s known that thefemale/maledifferences a r e due to relatively

greater pharyngeal length and more highly developed laryngeal cavities

TABLE I- C-11. b Average formant f r eq -~enc ies of voice c la s se s

=2 F3 F4

Ten 304 1969 2567 3105 A (0100) 9 13 3 7 Bar 2'78 1744 2482 2897 A(%) - 7 10 12 5 Bass 390 1557 2214 2'752

Ten 350 1942 2414 3061 A (70) 0 17 7 7 Bar 350 1662 2247 2373 A(%) - 2 B 10 4 Bass 356 1539 2041 2754

Ten 609 994 2576 2909 A( %) 15 5 7 2 Bar 530 944 2400 2349 A (%o) 5 5 I 13 Bass 503 900 2386 2527

Ten 401 724 2706 2989 a(%) 3 2 6 3 Bar 391 711 2554 2906 A!%) 7 -2 -2 -2 Bass 365 729 2605 2969

Ten 3 30 682 2548 2957 A! %) - 1 - 5 5 9 Bar 333 719 2420 2716 a!%) -4 - 4 -5 -2 Bass 348 749 2536 2784

A COMPARISON OF TENOR / BASS AND F E M A L E l W FORMANT FREQUENCY

DIFFERENCES

F2 F3

FORMANT FREQUENCIES Ihwels by IPA Symbols)

Fig. I-C-11.

S T L -QPSR 1/ 1976 28.

of adult males over those of females, i t i s likely that similar morpho-

logical distinctions exist between tenors and basses.

Since these vocal t ract differences a r e likely to exist between tenors

and basses, i t i s of interest to approximate what the differences in

vocal t ract length might be. Fant (1960) has shown that FZ in the vowel

[i ] i s clearly a half wave-length resonance of the back cavity. Further-

more, he has linked the dependence of the front cavity to F3 in the vowel

[ i 1. Utilizing Fant' s table of percentage dependence of given formant

frequencies on various cavities (Fant, 1960) hypothetical vocal tract

lengths of both tenors and basses can be approximated. Calculations I

I

show that overall vocal tract lengths for basses and tenors i s approxima-

tely 19 cm and 15, 5 cm, respectively, when singing the vowel [i 1.

Discussion ------ It is significant to note that these findings a r e not inconsistent with

the empirical observations of various laryngologists. Brodnitz (1961)

observed that, a s a rule, high voices a r e found in per sons with round

faces and short necks, while long faces and long narrow necks seem

frequent in singers with deep voices. Similar observations have been

made by Arnold (1965) and van Deinse (1974).

Concl usions ------- As a consequence of the various experiments presented in this paper

the following summary of conclusions seems to be in order:

1. Voice timbre type seems significantly dependent on formant I I

frequencies.

2. Voice classification may be estimated from voice timbre acoustic i cues - specifically the average of the four lowest formant fre- quencie s.

3. Measured formant frequencies from speech sounds seem indicative of voice classification.

4. Long-time-average- spectra seems to afford a good estimate of voice classification.

5. The mouth and pharynx lengths known to differentiate females and males may also differentiate tenors and basses.

Acknowledgments

The author i s deeply indebted to Dr. Johan Sundberg for his scholar-

ly guidance in the research and writing of this paper. The encourage-

ment of Professor Gunnar Fant, and the stimulating atmosphere provided

STL-QPSR 1/1976 29.

by the m e m b e r s of the Department of Speech Communication, KTH is

gratefully acknowledged. Appreciation i s a l so expressed to the singer-

subjects for their kind co-operation.

References

BRODNITZ, F. (1 9 6 1): "Vocal rehabilitation", American Academy of Ophthalmology and Otolaryngology.

COLEMAN, R. (1973): "A comparison of the contribution of two vocal character is t ics to the perception of maleness and femaleness in the voice", STL-QPSR 2-3, pp. 13-22.

Van DEINSE, J . , FRATEUR, L . , and KEIZER, J. (1974): "Problems of the singing voice", Folia Phoniatrica 26, pp. 428-434.

FANT, G. (1960): Acoustic Theory of Speech Production, ' s-Gravenhage, Mouton (2nd edition 1970).

FANT, G. (1966): "A note on vocal t r ac t s ize fac tors and non-uniform F-pat tern scalings", STL-QPSR 4, pp. 22-30.

FINN, J. (1974): A General Model for Multivariant Analysis, New York, Holt, Rinehart and Winston.

JANSSON, E. and SUNDBERG, J. ( ! 972): "Long-time-average-spectra applied to analysis of music", STL-QPSR 4, pp. 40-45.

LUCHSINGER, R. and ARNOLD, G. (1965): Voice, Speech, Language, Wadsworth P r e s s .

MI~LLER, G. (1973): "Statistische Langzeitanalyse von Sprechstimmen auf Frequenzgruppenbasis", Arch. Klin. exp. Ohr. -, Nas- u. Kehlk. Heilk. 204, pp. 151-160.

PETERSON, G.E. and BARNEY, H.L. (1952): "Control methods used in a study of vowels11, J.Acoust. Soc. Am. 24, pp. 175- 184.

SUNDBERG, J. ( 1973): "The source spectrum in professional singingf1, Folia Phoniatrica 25, pp. 71-90.

WOLFE, S. , STANLEY, D. , and SETTE, W. (19 35): "Quantitative studies on the singing voice", J. Acoust. Soc. Am. 6 , pp. 255-266. -

STL-QPSR 1/1976

SYMPOSIA PUBLICATIONS

I. The Proceedings of the Symposium on Auditory Analysis and P e r -

ception of Speech, which was held in Leningrad in August 1973, have

been published in two parts. One part, containing the neurophysiological

and psychoacoustic basis and auditory models, was published in a

special issue of Acustica, Vol. 31, No. 6, December 1974. I t can be

obtained from S. Hirzel Verlag, Stuttgart, BRD.

The second part of the Proceedings, including vowel perception,

feature theory and auditory segmentation, consonant perception, tem-

poral organization of connected speech and prosody, has been published

by Academic P re s s , London 1975 in a book Auditory Analysis and P e r -

ception of Speech , edited by G. Fant and M. A. A. Tatham.

2, The Proceedings of the Speech Communication Seminar, which was

held in Stockholm in August 1974, have been published by Almqvist &

Wiksell International, Stockholm 1975 and jointly by Halsted P r e s s , a

division of John Wiley and Sons, Inc. , New York. These Proceedings

a r e bound in four volumes, one for each session of the seminar, and

the title i s Speech Communication, Vols. 1-4. G. Fant has edited.

Vol. I: Speech Wave Processing and Transmission

Vol. 2: Speech Production and Synthesis by Rule

Vol. 3: Speech Perception and Automatic Recognition

Vol. 4: Speech and Hearing, Defects and Aids. Language Acquisition

These final volumes contain 28 articles not included in the preprint

version a s well a s comments on the sessions. The preprint version

was distributed to the participants.

The address of Almqvist & Wiksell International is:

P. 0. Box 62

S-101 20 Stockholm, Sweden

Here follows a list of articles in the Proceedings of the two symposia.

Symposium on Auditory Analysis and Perception of Speech, P a r t I pub-

lished in Acustica, December 1974. P a r t I1 published by Academic

P r e s s , London. Speech Communication Seminar (SCS-74) published

by Almqvist & Wiksell International, Stockholm with the title Speech

Communication.

P r o c . of the Symposium on Auditory Analysis and Percept ion of Speech, Leningrad, USSR, 21-24 Aug. 1973

G. Fant and M. A. A. Tatham (eds. ), Academic P r e s s , London 1975

Session 111. Vowel perception

R. Plomp: "Auditory analysis and t impre perception"

L. C. W. Pols: "Analysis and synthesis of speech using a broad- band spectra l representation"

E. G. Karnickaya, V. N . Mushnikov, N. A. Slepokurova. S. J a . Zhukov: "Auditory processing of steady- s ta te vowels"

R. Car l son . G. Fant , B. Grans t ram: "Two-formant models . pitch, and vowel perception'

R. C a r r t , R , Lancia: "Perception of vowels a~np l i t ude transic>nts1'

A. V. Baru: "Discrimination of synthesized vowels [ a 1 and j i 1 with varying p a r a m e t e r s (fundamental frequency. intensity, duration and number of formants) i n dog" (a summary)

W.A. Ainsworth: "Intrinsic and ex t r ins ic fac tors in vowt.1 judgments"

A. Q. Summerf ie ld , M. P. Haggard: "Vocal t r ac t normalization a s demonstra ted by react ion t imes"

K. Centmayer: "Interrelations of vowel perctaption and linguistic context"

E. ~ i s c h e r - ~ b r ~ e n s e n : "Percept ion of German and Danish vowels with specia l re fe rence to the German lax vowr.1~"

L. V. Bondarko, L.A. Verbitskaya: "Fac to r s undcrlying phonemic interpreta t ion of phonetically non-defined sounds"

Session IV. Fea tu re detection and auditory segmentation. Consonant perception

H. Fuj isaki , K. Nakamura, T. Imoto: "Auditory perception of duration of speech and non- speech st imuli"

L .A. Chistovich, N.A. Fyodorova, D.M. Lissenko. M . G . Zhukova: "Auditory segmentation of acoust ic flow and i t s pos sibli- role i n speech processing"

V. V. Lublinskaja. W. Mikiel: "Scaling of pitch intervals betwcxen vowellike sounds with constant and var iable ful~damentiil frequcnc)?'

T. F. M y e r s , M. G. Zhukova, L . A . Chistovich, V. N . Mushnikov: "Auditory segmentation and the method of dichotic stimulation"

J. M. Picket t , E. M. Danaher: "On discr iminat ion of formant t rans i - tions by persons with s eve re sensor i -neura l hear ing loss"

D. H. Klatt . S. R . Shattuck: "Percept ion of b r ie f st imuli that rc.st:mblc rapid formant transit ions"

K. N. Stevens: "The potential ro le of p roper ty detectors in thcx per- ception of consonants"

G. I. Tsemel : "Application i n speech recognit ion, somt. data o n auditory segmentation and speech wave paramc,tcr s ' perception '

K . Miyawaki, A. M. L iberman, 0. Fuj imura . W. Strange, J. J . Jenkins; "Cross-language study of tht. pc~rception of the F 3 cut.

f o r / r J v e r s u s 1 1 ] in speech- and nonsp<.erh-like pa t te rns i '

Table of contents

Vol. 1 SPEECH WAVE PROCESSIKG AND TRANSMISSION

FOREWORD by G. F a n t , A. R i s b e r g , and S. Fe l ice t t i , p. I.

CHAIRMAN'S COMMENTS ON SESSION 1 by J. N. Holrnes, pp. 11-IV

CHAIRMAN' S COMMENTS ON SESSION 1 by M. R. Schroeder . pp.

v-IX.

SOME COMMENTS ON LINEAR PREDICTIVE CODING by B. Gold.

pp. x-XXI.

ATAL. B. S . : "Towards de te rmin ing a r t i c u l a t o r posi t ions f r o m t h e

speech signal", pp. 1 - 9 .

XAKAJIh4A. T. ..OMURA, H.. TANAKA, K . . and ISHIZAKI. S . :

"Es t imat ion of vocal t r a c t a r e a functions by adaptive i n v e r s e

f i l t e r ing methods and identif icat ion of a r t i c u l a t o r y model",

pp. 11-20.

RICE D L : "Articulatory t rack ing of the acous t ic speech signal".

op. 21-26.

ATAL, B. S. and SCHROEDER, M . R. : "Recent advances i n pred ic t ive

coding - applicat ions to speech synthesis". pp. 27- 31.

L I W R A C E . L. A . : " L i n e a r estimation of speech signals". pp.

33-34 .

MAKHCIU L.. 3. : "Linear pred ic t ion vs . ana lys i s -by-synthes i s" .

pp. 35-43.

WAKITA. H. and GRAY J r . . A. H. : "Some theore t ica l cons idera t ions

for l inear predict ion of speech and applicat ions". pp. 45-50 .

ROZSYPAL, A. J. : "Computer supported gat ing of speech signals",

pp. 51-55 .

CARTIER. M . and GRAILLOT: P. : "Redur t ion and reconst i tut ion

of s p e c t r a l data", pp. i 7 - 6 4 .

JANSEK. L . P. C. : "Relat ions between s o m e of the p a r a m e t e r s of

vowel sounds". pp. 65-70 .

T a b l e of contents. Val. I

L-GLE, D. and PAULUS. E. : "Efficiency and l imitat ions of

l i n e a r t r a n s f o r m a t i o n s in digi tal speech t ransmiss ion" .

pp. 81-86 .

POLS. L. C. W.: "Intelligibility of speech resynthesized by us ing

a d imens iona l s p e c t r a l represen ta t ion" , pp. 87-95.

SUZUKI. H . . OOYAMA. G. . and KIDO, K. : "Analys i s -convrrs ion-

synthes i s s y s t e m for improving na tura lness and intel l igibi l i ty

of speech a t h i g h - p r e s s u r e hel ium g a s mixture", pp. 97-105.

SUZUKI. J . and NAKATSUI. M . : "Percept ion of speech u t t e r e d

u n d e r high ambient p r e s s u r e " , pp. 107-114.

* FLAWAGAN, 3. L. : "Digital coding of speech", pp. 115-117.

* GARDINI, B. and SERRA, A. : "Identification of speech p a r a m e t e r s

using a r e c u r s i v e method". pp. 119-128 .

* JAYANT. N. S.: "Spec t ra l a n a l y s i s of speech based on l i n e a r

del ta modulation", pp. 129- 137.

* LUKATE1.A. G . , TOMIC, T.. and DRAJIC, D. : "Efficiency

a n d a c c u r a c y in vocal pi trh de te rmina t ion by some m e t h o d s

us ing digital computer" . pp. 139-147 .

* HOLLIEN. H. , FEINSTEIIG. S. H. . ROTHMAN. H. . and HOI.LIEN. P. :

"Underwater SONAR s y s t e m s (passive) in humans". pp.

149-157.

* t h e s e p a p e r s w e r e not published in the p r e p r i n t vers ion

McLARNOK, E . . HOLMES. J . K . . and JUDD. M . W . : " E x p e r i m e n t s

with a v a r i a b l e - f r a m e - r a t e coding s r h e m e applied to f o r m a n t

synthes izer control s ignals". pp. 71-79 .

Table of contents

Vol. 2 SPEECH PRODUCTION AND SYNTHESIS BY RULES

CHAIRMAN'S COMMENTS ON SESSION 2 by J. L . F lanagan , pp. 1-11.

CHAIRMAN'S COMMENTS ON SESSION 2 by B. Lindblom. pp. 111-XIX

MacNEILAGE, P. F.. HANSON, R. J. . a n d SUSSMAN. H.M. :

"Studies of s ing le m o t o r unit ac t iv i ty i n speech musculature".

p p . 1-7.

F R I T Z E L L , B. . KOTBY . M. N. . a n d MOLLER. B. : "Var la t ions

i n the ac t iv i ty of the palatopharyngeus m u s c l e d u r i n g speech".

pp. 9-14.

IMAIZUMI. S. and HIKI, S. : "Ext rac t ion of m o t o r command f r o m

e lec t romyographic data", pp. 15-21.

MANSELL. P. : "Some EMG data o n the a r t icu la t ion of [ f ] and [ p ] ". pp. 23-32.

FUJIMURA. 0. : "Articulatory observa t ions and i t s l inguis t ic

implicat ions", pp. 33-43.

KAKITA. Y. alsd HIJSI. S. : "A study of l a r y n g e a l cont ro l f o r vo ice

pitch based o n ana tomica l model", pp. 45-54.

LUBKER. J. , McALLISTER. R. . a n d CARLSON, J. : "Labial

co-ar t icu la t ion in Swedish:, A p r e l i m i n a r y repor t" , pp. 55-63.

OHALA, J. J. : ''A m a t h e m a t i c a l model of speech aerodynamic.".

pp. 65-72. , .

LANCE. D. and van d e r CIET. G. : ''A c o m p u t e r on-l ine method f o r

m e a s u r i n g a r t i c u l a t o r y movements". pp. 73-77.

HAMLET, S. and STONE. M. : "Reorganizat ion of s p e e c h motor

p a t t e r n s following prosthodontic changes i n o r a l morphology".

pp. 79-86.

LINDBLOM, B. E. F . , PAULI. S . , and SUNDBERC, J.: "Modeling

coar t icu la t ion in ap ica l stops". pp. 87-94.

KOHLER. K . and HARDCASTLE, W.: "The instabil i ty of f ina l

a l v e o l a r s i n Engl i sh and German: P r o p o s a l f o r a n i n s t r u m e n t a l

investigation". pp. 95-98.

LC)FQVIST. A . : "Var ia t ions in subglottal p r e s r u r e dur ing s t o p

production". pp. 99- 104.

BENGUEREL. A-P. : "Nasa l a i r f low p a t t e r n s and v e l a r coar t icu la t ion

i n French". pp. 105-112.

v a n d a r GIET. G, and LANCE. D.: "Influence of t ime on f o r m a n t

f r e q u e n c i e s e s t i m a t e d by a t ime dependent horn equation".

pp. 113-119.

FANT. G. a n d PAULI. S.: "Spatial c h a r a c t e r i s t i c s of vocal t r a c t

r e s o n a n c e modes". pp. 121- 132.

OHMAN. S. E . G. and ZETTERLUND. S. : "On s y m m e t r y in t h e

vocal t rac t" , pp. 133-138.

FANT. G. . STALHAMMAR. L'. . and KARLSSON, I. : "Swedish

vowels i n speech m a t e r i a l of var ious complexity". ~ p . 139-147

KORD. L . : "Vowel reduction-central izat ion o r contextual a s s i m i l a -

t ion?" . pp. 149-154.

O L L E N W R F F . F. and KAPLAN, G. : "The effect of a c le f t pa la te

o n the acous t ica l s p e c t r u m of vowels". pp. 155-162.

IIVONEN. A . : "On the re la t ionsh ips between the m a l n components

in the cont ro l of speech". pp. 163-1171.

LAWREXCE. W. : "The phoneme. the syl lable. and the p a r a m e t e r

t rack". pp. 173-t78.

LIIV. G. a n d REMMEL, M. : "Es t imate of the d i s t inc t ive p a r a m e t e r s

i n the domain of t iming , fundamental frequency and in tens i ty

with impl ica t ions for modeling of a quanti tat ive sys tem", pp.

179-185.

E E K , A. and REMMEL.. M. : "Context, contac ts and durat ion: Two

r e s u l t s concerning t e m p o r a l organizat ion". pp. 187-192.

POTAPOVA. R. K. : "Some t iming c h a r a c t e r i s t i c s of spoken R u s s i a n

(data r u l e s ana lys i s - synthes i s -ana lys i s )" , pp. 193-196.

BLOCHINA. L. P.: "On mode:ling f requency f e a t u r e s in a phrase" .

pp. 197-2F4.

GAY. T . and L'SHIJIMA. T. : "Effect of s p e a k i ~ g r a t e on s top

consonant-vowel art iculat ion". pp. 205-208.

CARLSON. R . . ERIKSON. Y . . GRANSTROM. B . . LINDBLOM, B..

and R A P P . K . : "Neutral and emphatic s t r e s s pa t te rns in

Swedish". pp. 209-218.

HIRT, ;. and OIZUMI. J. : "Speech synthemim by rule from neuro-

physiological parameter", pp. 219-225.

SCULLY, C.: "A synthesiaer study of aerodynamic fac tors in

speech segment durations". pp. 227-234.

ROTHENBERG. M., CARLSON, R.. GRANSTRC)M, B.. and

LINWVIST-GAUFFIN. J. : "A three-parameter voice source

for speech synthesis", pp 235-243.

CARLSON, R. and GRANSTRCh4 , B. : "A phoneticdly orientecl

programming language for ru le description of speech", pp.

245-253.

OLIVE, J. P. : "Speech synthesis by rule", pp. 255-260.

NAKATA. K. and ICHXAWA, A. : "Speech synthesis for a n

unlimited vocrbulary". pp. 261-266.

ENDRES, W. and GROSSMANN. E. : "Manipulation of the t ime

functions of vowels for reducing the number of elements

needed for speech synthesis", pp. 267-275.

KIM. B. and F'UJISAKI, H.: "Measurement of mandibular control

in vowels and i t s relevance t o the articulatory description of

the vowel syrterno of Korean and Japanese", pp. 277-284.

JOSPA, P. : "On the use of cr i t ica l formant frequencies in designing

an articulatory model". pp. 285-292. ," ,

ONDRACKOVA, J. and MACHEK. J.: "The production of vowels

i n the s t r e~aed /uns t r eased position", pp. 293-298.

HARRIS, K. S. : "Mechanisms of duration change", pp. 299-305.

MEZZALAMA. M. and RUSCONI, E. : "A general system for

synthesizing speech", pp. 307-314.

MEZWLAMA. M. and RUSCONI, E.: "Intonation in speech synthesis:

A preliminary a d d y for the Italian language". pp. 315-325.

SELL. K. : "A digital speech synthesiaer fo r simultaneous voice

output on severa l independent ch8nnels". pp. 327-331.

' these *per# were not pub1ish.d ia the pr.)rh mrlon

Table of contents

Vo13 SPEECH PERCEPTION AND AUTOMATIC RECOGNITION

CHAIRMAN' S COMMENTS ON SESSION 3 by M. Haggard, pp. I-VIII.

CHAIRMAN'S COMMENTS ON SESSION 3 by D.H. Klatt , pp. IX-XXVIII.

SUSSMAN, H. M. and MacNEILACE, P. F. : "Some studies of

hemispheric specialization for speech production", pp. 1-8.

BERLIN, C. I. . CULLEN J r . . J. K . , LOWE-BELL, S. S. . and

BERLIN, H. L. : "Speech perception af ter hemispherectomy

and temporal lobectomy", pp. 9- 15.

CRYSTAL, T . H. and HOUSE, A. S. : "Local signal levels and dif-

ferential performance in dichotic listening", pp. 17-20.

PORTER. R. J.: "The dichotic lag effect: Implicationr for the central

processing of speech". pp. 21-30.

CULLEN, J r . , J .K. , THOMPSON, C. L.. HUGHES, L. F.. and

BERLIN, C. I:: "Information additivity i n dichotic stop-vowel

perception tasks". pp. 3 1-37.

HAGGARD, M.: "Pat tern masking and the speech perception

procesr", pp. 39-45.

BAILEY. P. J . : "Perceptual adaptation for acoustical fea tures in

speech", pp. 47-53.

SINGH. S. : "A s tep towardr a theory of speech perception''. pp. 55-66.

DRAPER. G. and HAGGARD. M. : "Facts and ar t i fac ts i n feature

interdependence", pp. 67-75.

SUMMERFIELD, A. Q. : "Procesring of cues and contexts i n the

, perception of voicing contrarts". pp. 77-86.

SERNICLAES. W. : "Perceptual processing of acoustic co r re l a t e s

of the voicing feature", pp. 87-94.

RAO. P. V. S. : "On s top consonants", pp. 95-102.

SOTSCHECK. J. : "Stop consonants identification of bandlimited

transmiroton ryr temr" , pp. 103-109.

COHEN, A. : "Formant discrimination i n the auditory system", pp.

111-116.

FANT, G. . CARLSON. R. , and GRANSTROM, B. : "The [ e ] - [ 4 1 ambiguity", pp. 117-121.

AINSWORTH. W.A. : "Influence of fundamental frequency on perceived

vowel boundaries in English", pp. 123-129.

GERBER, S. E. : "Categorical perception of segmented diphthongs",

pp. 131-136.

TSEMEL' . G. I. and KRINOV, S. N. : "On the initial and final t r an r i -

tional segments of vowels". pp. 137-140.

TAKEUCHI. S. . KASLIYA. H. . and KIDO. K.: "An active model

for extraction of nasali ty and i t s perceptual evaluation". pp.

141-148.

NOOTEBOOM. S. G. : "Contextual variation and the perception of

phonemic vowel length", pp. 149- 153.

GARNES, S. : "Implications of the perception of duration", pp. 155-164.

CARTERETTE, E. C. and HUBBARD JONES, M. : "On the s ta t i s t ics

of spoken American English", pp. 165-173.

JASSEM, W. : "Computer-aided recognition of Polish vowels in

continuous speech", pp. 175-181.

LIENARD, J. S. , MMUKA, M . . MARIANI. J . 3. . and SAPALY. J.: "Real-time segmentation of speech". pp. 183-187.

BUISSON, L.. MERCIER. G.. CRESSER. J.Y.. QUERRE. M..

and VIVES. R. : "Phonetic decoding for automatic recognition

of wordr", pp. 189-196.

HATON. J-P. : "Current work in segmentation and analytic81

recognition of speech", pp. 197-200.

HESS. W. 3. : "A pitch-aynchronour, digital feature extraction

sys tem for phonemic recognition of speech", pp. 201-213.

PAULUS, E . , SCHRAG, R. . and SCHOTOLA, Th. : "Advancer

towards o four-stage system for automatic rpeech recognition".

pp. 215-223.

WHITE. G. M. : "Simple technique. for trannforming rpeech to

quaot-phoneme stringr", pp. 225-231.

Table of contents

Vol. 4 SPEECH AND HEARING, DEFECTS AND AIDS.

LANGUAGE ACQUISITION

CHAIRMAN'S COMMENTS ON SESSION 4 by I. 3. Hirsh , pp. I-IV

CHAIRMAN'S COMMENTS ON SESSION 4 by I. Lehiste, p. V.

MAVILYA. M. : "Aural-oral communication begins a t birth". pp. 1-5.

SIMOX, C. : "Some aspec t s of the development of speech production

and perception in children". pp. 7-14.

BRYDES. 3. D. and GRAY. 9. B. : "Programmed conditioning for language", pp. 15-19,

JVBAN. P . . GUYARD. H . . BIAQUET, Y. hi. . r a d DUVAL, A.: "Towards computer-aided simulator of phonological aphasia". pp. 21-23.

KAMARA. C .A . , KAMARA, A. S., and SINGH. S.: " 'Featuregram' analysis of art iculation performance in var ious hearing and

speech pathologies", pp. 25-34.

CRICHTON. R. G. and FALLSIDE. F. : "The development of a deaf

speech training a id using l inear prediction", pp. 35-40.

EDMONDSON. W. H. : "Pre l iminary experiments with a new vibro-

tactil; speech training aid for the deaf". pp. 41-47.

EDMONDSON, W. H. : "Amplitude insensit ive filtering". pp. 49-51.

JISUALM. G. : "A communication aid with speech output and built

in grammar" . pp. 53-56.

KORSAN-BENGTSEN, M. : "Low-redundant speech t e s t s - A method

to evaluate d is turbances in the cent ra l hearing mechanisms",

pp. 57-62.

DANAHER. E. M. and PICKETT, J. M. : "Temporal masking i n s e n ~ o r i n e u r a l subjects and i t s effect on syllable discrimination".

pp. 63-68.

MARTONY. J.. and AGELFORS, E.: "Two psychoacoustic t e s t s with

severe ly hard of hearing children". pp. 69-76.

CONSTAM, A. G. : "Comments on clinically usable method8 f o r hearing

aid fitting. A new low peak narrowband noise genera tor f o r

rea l i s t ic measurement of hearing aid amplification". pp. 77-78.

HIRSH. I. J. : "Recent development8 in p~ychoacoust ic tools and

hearing aids". pp. 79-82.

STEWART, R. B. : "A linguistic look a t lipreading". pp. 83-91.

MARTONY. J . . AGELFORS. E . . BLOMBERG. M. . BOBERC. G..

ELENNS, K. , RISBERG. A . , SPENS. K-E., and OSTER. A-M.:

"Experiments with electronic lipreading aids". pp. 93-96.

MILLER, J. D. : "Pre l iminary research with a three-channel vibro-

tacti le speech-reception a id for the deaf", pp. 97-103.

PICKETT, I . M . . GENGEL. R. W . , and QUINN. R. : "Research with

the Upton eyeglass speechreader", pp. 105-109.

TRAUNMULLER, H. : "A visual lipreading aid", pp. 111- 120.

KARLSSON, I. and STALHAMMAR, U. : "Spectrographic r eg i r t r a t i ons

a s a id s for pronunciation control". pp. 121- 127.

Q LEVITT. H. . SMITH, C . R . . and STROMBERG, H. : "Acoustic.

a r t icula tory and perceptual character is t ics of the speech of

deaf children", pp. 129- 139.

* SURLAN, A. : "A new audio-visual rys tem fo r the communictt ion of

the deaf andmute" , pp. 141-145.

* MARTONY. J . , AGELFORS. E . , BLOMBERG, M. . BOBERC. G. . ELENIUS. K.. RISBERC. A.. SPENS. K-E., and OSTER. A-M.: "Fur the r experiments with electronic l ipreading aid.", pp.

147-156.

*these paper. we re not published la tho proprlnt vroioa