Dept. for Speech, Music and Hearing

Quarterly Progress andStatus Report

Acoustic comparison of voiceuse in solo and choir singingRossing, T. and Sundberg, J. and Ternstrom,

S.

journal: STL-QPSRvolume: 25number: 1year: 1984pages: 030-043

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

LEVEL OF ACCOMPANIMENT (dB)

CHOIR

90 -

*

60 70 80 90 LEVEL OF CHOIR (dB)

Fig. 1. Canparison of sung level to the reference level heard by the subject. Each data point represents one sung note. In solo the reference was the piano ac-t, and in choir the reference

, was the sound heard at a seat in the bass section. In the solo mde, the sing- er always sang at a higher level than the reference, whereas i? the choir mode this was true only at low reference levels.

BE

..

' .-3

1 _ . . - loud so10 , --- so f t solo

...... loud choir --- soft choir i

BAND FREQUENCY (Hz) Fig. 2. Long-tenn-average spectra of two solo passages and tim choir

passages by a professional bass/baritone singer. Spectra are ccanputed in 1/3 octave bands, and the levels of the curves have been adjusted to be the same at 500 Hz. -

Table I summarizes some of the information from the LTAS of the e igh t subjects. By and large, LSo0 ( the leve l i n the 500 Hz band) corresponds to the f i r s t formant, L200 Hz (the level i n the 200-Hz barrd) corresponds to the fundamental, and kF (the maximum band level i n the 2-4 kHz range) corresponds t o the "singer's formant". The leve ls ( i n dB) are averages of 3 solo passages and 7 choir passages, respectively, sung a t different dynamic levels. In a l l subjects except two , the "sing- er's formant" was more prominently used i n so lo singing than i n choir singing. A l l eight subjects produced Inore energy in the lowest frequency range, on the average, i n choir singing than i n so lo singing. Subjects J B and WW had r e l a t i ve ly weak "singer's formants" i n both modes, and subject JS had a strong "singer's formant" i n both modes. Subject TR, a tenor, produced the leas t fundamental power i n singing the bass part in the choir setting.

Table I. Average differences i n LTAS level (L) between different frequency bands in choir and solo singing.

Subject Difference

Professionals : *

BE 0.3 7.1 m -5.7 7.3 LW 1.0 7.3

A chart of sound pressure level for one solo passage arrd one choir passage sung by subject BE is shown i n Fig. 3. Also shown is the sound leve l i n the 2-4 kHz range. In the so lo passage, there is a trend t o a constantly high amount of power i n the range of the "singer's formant", w h i l e i n the choir passage the power i n this range is mostly lower and considerably more variable. I t seems as i f the "singer's formant" is used, consciously o r unconsciously, t o emphasize o r decorate ce r t a in notes. In many syllables, particulary i n choir singing, a delayed "sing- er's formant" appears. The strategies of the different singers varied considerably i n this respect.

According to Fig. 2 there is more energy i n the "singer's formant" and less energy i n the lowest range of the spectrum i n so lo singing. Such spectral differences are typically associated with differences in vocal loudness. As vocal loudness is raised, the higher spectral com- ponents gain more i n amplitude than the lower spectral components. Thus, the LTAS differences between the solo and choir singing might suggest that vocal loudness tended to be higher i n solo singing.

I - solo I t --- choir 1 1 I I I I

1 I

100 200 500 1000 2000 5000 10000

BAND FREQUENCY (Hz)

0 1 2 3 4 FREQUENCY (kHz

Fig. 4. TLAS which have the same Fig. 5. Spectrum of a vuwel sound level in the 500-Hz band. Be- /o/ before and after in- 500 Hz the levels are greater verse filtering (solid in choir singing, whereas a- and dashed curve, respect- W e 500 HZ, they are greater L . ively) . The six formant in solo singing. This is con- frequencies are shown be- sistent with the LTAS fram a law by arraws. differentsingershowninFig.2.

FREQUENCY (kHz ) FREQUENCY (kHz)

Fig. 6. Source spectra of t m matching vow1 pairs (i . e . , the same vuwel , the same pitch, and the sarne sound pressure level). In the first one, the voice source spectrum in choir (dotted curve) shows more p e r at the fundamntal frequency, whereas the voice source spectrum in solo (solid curve) is greater in the range of the "singer's formant". In the other pair, the voice source spectra are more nearly the same insoloandchoir.

too, pa r t i cu l a r ly i n t he temporal aspects. For example, we would expect a t tacks and releases with a gl id ing p i t ch t o occur much more frequently i n solo singing, an3 such glides may help the solo singer's voice to stand out more clearly against a loud accompanying sound.

In devising the present experiment, ca re was taken t o o f f e r t he subjects as r e a l i s t i c experimental conditions as possible. A s it w a s necessary to record the subject's own phonation separately, the sound of the accompanying music had to be presented aver earphones that t ightly sealed the subject's ea r canals. This had the e f f e c t t h a t the subject heard the low frequency components of his own voice a t too high a sound level (Tbnndorf, 1972). Some subjects noticed this and occasionally had some m i n o r diff icult ies of synchronizing their singing with the accompa- niment. However, the e f f e c t of t h i s on the r e su l t s , i f any, would be s i m i l a r for solo and choir singing.

Goodwin (1980) found t h a t female so lo s ingers sang a t a higher sound pressure leve l when singing i n a so lo mode than when t rying to achieve a "choral blend". In normal speech, such sound pressure leve l differences lead t o sound leve l differences i n t he various spec t r a l bands (Gauffin & Sundberg, 1980). For instance, an increase i n vocal loudness is normally associated with a greater amplitude increase i n the spectrum partials a t high frequencies than a t l ow frequencies. Goodwin substituted a rated value of perceived loudness for the sound pressure level when he compared spectra from the two singing modes . Therefore, it is not c l ea r from h i s r e su l t s how much of h i s spectrum differences a r e due t o t he difference i n sound pressure level . Our r e s u l t s show that some spectral differences remain between solo and choir singing, even when vowels a r e sung a t the same sound pressure level. The main differences which we observed are (1) a stronger amplitude of the source spectrum fundamental and (2) a lower amplitude in the frequency range of the "singer's formant" in choir singing.

The spectral differences revealed by LTAS w a r to agree with the r e su l t s from inverse f i l t e r i n g . The higher amplitude of the "singer's formant" i n the solo singing seems to be an effect of formant frequency tuning, particularly the frequency spacing of the third, fourth and and f i f t h formants, which reflects articulatory differences between the two modes of singing. The higher amplitude of the LTAS in the low frequency range i n choir singing is compatible with the trend both t o a lower f i r s t formant frequency and t o a higher amplitude of the voice source fundamental. According to previous investigations, this l a s t mentioned voice charac te r i s t i c corresponds t o a higher peak amplitude of t he tramglottal a i r flow waveform, which can be achieved by reducing the degree of g l o t t a l adduction a c t i v i t y and by lowering the subglo t t i c pressure (~undberg & Gauffin, 1979; Gauffin & Sundberg, 1980). The corresponding type of phonation has been called "flow phonation". l?ha,

it s e e m s t h a t some o f t h e spec t ra l differences betweensoloand choir singing also reflect differences i n glot ta l adjustment.

What purpose could these differences serve? In choral compositions the bass singers are usually responsible for carrying the important bass

I n some cases t h e e f f e c t is increased by a h igher amplitude of t h e g l o t t a l voice source overtones i n the corresponding frequency range. In s o m e s ingers a s t ronger fundamental is a t t a i n e d by a lowering of the f i r s t formant frequency.

Thus, solo and cho i r s inging s e e m to show d i f fe rences i n both ar t iculat ion and phonation.

W e wish t o thank t h e s i n g e r sub jec t s who p a r t i c i p a t e d i n these experiments and a l s o Lennart Nord and Johan L i l j enc ran t s f o r t h e i r ass is tance . The p r o j e c t received support from t h e National Science Foundation (USA) through t h e U.S.-Scandinavia Cooperative Science Pro-

gram and from the Swedish Research Council fo r the H u m a n i t i e s and Social Sciences, fo r which we a re grateful.

I

i t 1

Wferences Cleveland, T.F. (1977): "Acoustic properties of voice timbre types and t h e i r inf luence on voice c l a s s i f i c a t i o n " , J.Acoust.Soc.Am. - 61, 1622- 1629.

Gauff i n J. & Sundberg, J. (1980): "Data on t h e g l o t t a l voice source behavior i n vowel production", STLAPSR 2-3/1980, 61-70.

Goodwin, A. (1980) : "Acoustic s tudy of indiv idual voices i n chora l blend", J. of Research i n Singing - 3:2, 25-36.

Imttermos er, W. (1976177) : "Frequenzschwankungen bei musikalischen KlZngen", k u s t i c a - 36, 138-146.

Sacerdote, G.G. (1969) : "Researches on t h e s inging voice", Acustica - 7, 61-68

Sundberg, J. (1974): "Acoustic i n t e r p r e t a t i o n of t h e 'singer's for- mant'", J.Acous t.Soc.Am. - 55, 838-844.

Sundberg, J. (1977) : "Singing and timbre", i n Music m m Acoustics, 57- 81; Publicat ions issued by t h e Royal Swedish Academy of Music No. 17; this reference contains a sound i l lus t ra t ion of the effec t of the "sing- er's formant" on the audibi l i ty of a voice presented against the back- ground of loud noise.

Sundberg, J. 6( Gauffin, J. (1979) : "Waveform and spectrum of the g l o t t a l voice source", pp. 301-320 i n (B.Lindblam & ~.6hman, eds.), Frontiers of Speech Communication Research, Academic Press London. !-, ,. Ternstrom, S. & Sundberg, J. (1982): "Acoustical f a c t o r s r e l a t e d to pi tch precision i n choir singing", STGQPSR 2-3/1982, 76-90.

Ternstrom, S. & Sundberg, J. (1983) : "How loudly should you hear your colleagues and yourself? A s tudy of SPL wi th in choirs" , STL-QPSR 411983, 16-26.

STL-QPSR 1 / 1 9 84

Ternstrom, S., Sundberg, Jet & Collden, A. (1983) : "Articulatory per- turbation of pitch in singers deprived of auditory feedback8', !~TG.QPSR 4/1983, 144-155.

Tonndorf, J. (1972): "Bone conduction", pp. 195-237 i n Foundations of Modern Auditory Theory, (JUX'obias, ed.) , Academic Press New Y o r k