Acoustical Study of Classical Peking Opera Singing

Acoustical Study of Classical Peking Opera Singing *Johan Sundberg, †Lide Gu, †Qiang Huang, and †Ping Huang, *Stockholm, Sweden, yBeijing, China Summary: Acoustic characteristics of classical opera singing differ considerably between the Western and the Chinese cultures. Singers in the classical Peking opera tradition specialize on one out of a limited number of standard roles. Audio and electroglottograph signals were recorded for four performers of the Old Man role and three performers of the Colorful Face role. Recordings were made of the singers’ speech and when they sang recitatives and songs from their roles. Sound pressure level, fundamental frequency, and spectrum characteristics were analyzed. Histograms showing the distribution of fundamental frequency showed marked peaks for the songs, suggesting a scale tone structure. Some of the intervals between these peaks were similar to those used in Western music. Vibrato rate was about 3.5 Hz, that is, considerably slower than in Western classical singing. Spectra of vibrato-free tones contained unbroken series of harmonic partials sometimes reaching up to 17 000 Hz. Long-term-average spectrum (LTAS) curves showed no trace of a singer’s formant cluster. However, the Colorful Face role singers’ LTAS showed a marked peak near 3300 Hz, somewhat similar to that found in Western pop music singers. The mean LTAS spectrum slope between 700 and 6000 Hz decreased by about 0.2 dB/octave per dB of equivalent sound level. Key Words: Singing–Classical Peking opera–Long-term-average spectrum–Scale tone intervals–Singer’s formant cluster–Vibrato–Spectrum slope.

INTRODUCTION Like Chinese culture in general, the classical Chinese opera art has a long history. It reflects ancient Chinese thought, culture, religion, and other ideological elements and uses song, dance, speech, recitative, preaching, rituals, burlesque, and acrobatics, as performing vehicles. Jing Ju (Peking opera) incorporated the characteristics of ‘‘comprehensive performing art,’’ ‘‘virtual imitation,’’ and ‘‘programmed acting measurements.’’ It was created in 1790 in connection with the celebration of the Emperor’s birthday. It was built mainly on the essence of the Hui Ju and the Han Ju, which were two opera traditions developed in the An Hui and Wu Bei provinces. It began to form and develop rapidly, stimulated by the Emperor’s preference and other factors such as the use of a number of popular stars, especially for the male roles. Soon, the Peking opera was unanimously recognized as the main opera among the more than 300 types of concurrent Chinese operas. Its performance forms and role classification, including the voice specifications of these roles, were widely borrowed also by other types of Chinese operas. In the Peking opera, the performers can be divided into four different types of voices: 1. Sheng, including the two male roles of Lao Sheng and Xiao Sheng. The former is always masked with a false beard and plays the roles of a middle-aged or an elderly gentleman, speaking and singing with a ‘‘big’’ or ‘‘true’’ male voice. Xiao Sheng plays young male roles,

Accepted for publication January 7, 2011. From the *Department of Speech Music Hearing, School of Computer Science and Communication, KTH, Stockholm, Sweden; and the ySinging Voice Research Institute of China Conservatory, Beijing, China. Address correspondence and reprint requests to Johan Sundberg, Department of Speech Music Hearing, School of Computer Science and Communication, KTH, Stockholm SE-10044, Sweden. E-mail: [email protected] Journal of Voice, Vol. 26, No. 2, pp. 137-143 0892-1997/$36.00 Ó 2012 The Voice Foundation doi:10.1016/j.jvoice.2011.01.001

carries no beard, and speaks and sings with ‘‘small’’ but ‘‘masculine,’’ strong, falsetto voice. 2. Dan, includes the Hua Dan, Qing Yi, and Lao Dan roles. In general, Hua Dan plays young girl’s role, whereas Qing Yi plays the role of a middle-aged woman. Both speak and sing with ‘‘small’’ (falsetto) voices. Lao Dan plays the role of an old lady, speaks and sings with ‘‘big’’ normal speaking voice. 3. Jing, also called Da Hua Lian, ‘‘colourful face,’’ or if verbatim translated ‘‘big flower face’’, an allusion to how their faces are being painted; speaks and sings with a special type of voice. 4. Chou, also named Xiao Hua Lian, ‘‘small flower face’’ if translated verbatim, plays the role of a clown, and speaks and sings with register shifts between falsetto and modal. The voice timbres used in these roles differ dramatically from those used in the Western operas, but few attempts have been made to describe in scientific terms their acoustic characteristics. The study by Wang1 is an exception. This author found that the Peking opera singers sang with an elevated position of the larynx and that, still, their voices contained a ‘‘singing formant.’’ This was a surprising observation, as coauthor J.S.2 had demonstrated that a low larynx position promoted the formation of this spectrum envelope peak, characteristic of Western male opera singer voices. The peak can be explained as the result of a clustering of formants 3, 4, and 5, a phenomenon that can be expected to occur if the cross-sectional area of the pharynx is wide as compared with the entrance to the larynx tube.2 This condition can typically be fulfilled if the larynx is lowered. In other words, Wang’s finding seemed to be at variance with that interpretation of the singer’s formant. The findings by J.S. were based on analyses of Western opera singers, so it seemed worthwhile to attempt to replicate the findings by Wang. The present investigation is focused on two Peking opera roles, the elderly gentleman role Lao Sheng and the Colorful Face role Da Hua Lian. The purpose was to identify their acoustic characteristics.

138

RESULTS The loudness relations between the different voice samples and the two roles are illustrated in Figure 1, showing the distribution of SPL values at 0.3 m for the three conditions Speech, Recitative, and Singing. The averages across singers for the roles are listed in Table 1. In speech, the distribution of SPL values appears symmetrical around the mean. On average, the speech samples were about 12 dB softer than the recitative samples for both roles. The values for the Song tended to be a few dB higher than those for the Recitative for both roles. The mean F0 in the Recitative and the Song was markedly higher than that of the Speech samples (Table 2). On average, the difference between mean F0 in Speech and mean F0 in Recitative and Song was larger in the Colorful Face role than in the Old Man role. In the latter case, the mean F0 in both Recitative and Song mostly exceeded the mean F0 in speech by more than

0,04 Singing Colourful face Recitative Colourful face

Occurrence [%]

METHOD Seven male professional representatives of the classical Peking opera style of singing, all from the China Theatre University, volunteered as subjects. Four of them had specialized for singing the Old Man role, and three for singing the Colorful Face role. Each of the singers was asked to read a text lasting for about 40 seconds. As the phonetic content of the text of this duration is irrelevant to the appearance of a long-termaverage spectrum (LTAS), the subjects were allowed to select different texts. In addition, they sang a recitative and one or two songs from the Peking opera repertoire. Audio was picked up by (1) a DPA 4061 (Longmont, CO) headworn omnidirectional miniature condenser microphone mounted off axis at a distance of 30 cm. An electroglottograph signal was recorded by means of a KAYPENTAX (Lincoln Park, NJ) Model 6103 device. These signals were digitized at a rate of 44.1 kHz (bit depth 16, pulse code modulation coding) and recorded on dual channel wav files into a KAYPENTAX Computerized Speech Lab machine, model 4500. Sound pressure level (SPL) calibration was carried out by recording a sustained tone, the SPL of which was measured at the recording microphone by means of a TES-52 Sound Level Meter (TES Electrical Electronic Corp., Taiwan, ROC). This SPL value was announced in the recording file together with the microphone distance. The recordings were made in an ordinary room, with a volume of about 60 m3, reverberation time of about 0.5 second. For analysis, the wav files were converted into files of the smp format, thus allowing analysis by means of the Soundswell program (Soundswell Core Signal Workstation 4.0, Saven Hitech, T€aby, Sweden). The following parameters were measured: fundamental frequency (F0), tracked by means of the Corr module; equivalent sound level (Leq) determined by means of the Histogram module, and LTAS from 0 to 16 000 Hz determined by means of the Sect module, using an analysis bandwidth of 200 Hz. In addition, again using the Histogram module, the distributions of F0 and SPL values in the speech, recitative, and aria samples were analyzed.

Journal of Voice, Vol. 26, No. 2, 2012

Singing Old man Recitative Old man

0,03

Speech

0,02

0,01

0,00 60

70

80

90

100

110

120

SPL @ 0.3 m [dB]

FIGURE 1. Distributions of SPL values for the indicated types of voices.

one octave. As indicated by the standard deviations, the variability of F0 was greater in the Recitative than in the Song for both roles. Different music cultures use different tuning systems for their scales. Hence, it seemed relevant to examine the musical intervals between the scale tones. Such intervals can be measured from F0 histograms. Typical examples of distributions of F0 values for Speech, Recitative, and a Song in the Old Man and Colorful Face roles are shown in Figure 2. The distribution for Speech shows a few weakly marked peaks at 2.5 and +2.5 semitones below and above 220 Hz and also one at 4.5 semitones above 220 Hz. In other words, the intervals between the lowest and the two higher peaks were five and seven semitones. Unlike the F0 histograms for the Recitative samples, those for the Song samples showed several distinct peaks. The musical intervals between these peaks were measured and expressed in semitones. The results are listed in Table 3. As can be seen in Table 3, several examples were observed that agreed within 10 cent with intervals commonly used in Western music cultures: major second, minor and major third, pure fourth and fifth, and pure octave. However, many peaks were also separated by intervals lying almost halfway between those used in traditional Western music performances. Figure 3 shows a histogram of all intervals observed within the songs between all combinations of histogram peaks. Major second, minor third, pure fourth, and pure fifth occurred more than four times in the histogram material. The singers sang certain tones with vibrato. Figure 4 shows two typical examples. In the beginning of such tones, the vibrato rate was in the vicinity of 3 Hz, which is about 2 Hz slower than in Western classical singing, and it sometimes was TABLE 1. SPL Averaged Across Subjects of for the Indicated Voice Samples Type Speech Recitative Old Man Singing Old Man Recitative Colorful Face Singing Colorful Face

Average (dB SPL at 0.3 m) 79.1 91.7 94.0 92.6 94.1

Johan Sundberg, et al

139

Acoustical Study of Classical Peking Opera Singing

TABLE 2. Mean and SD of F0 Observed for the Indicated Conditions Speech

Recitative

Song

MVF0 (Hz)

SDF0 (Hz)

MVF0 (Hz)

SDF0 (Hz)

Re F0Speech (Semitones)

MVF0 (Hz)

SDF0 (Hz)

Re F0Speech (Semitones)

Old Man role Singer 1 Singer 2 Singer 3 Singer 4

141.3 180.2 137.8 191.3

22.7 50.9 21.2 40.1

244.8 271.2 299.1 272.6

79.3 100.5 91.9 93.9

9.5 7.1 13.4 6.1

233.8 285.9 295.9 313

58.8 104.8 62 65.9

8.7 8.0 13.2 8.5

Mean SD

162.7 27.1

8.9

282.2 34.1

10.6 17.2 12.5

384.7 340.8 391.2

13.2

372.2 27.4

271.9 22.2

Colorful Face role Singer 5 142.1 Singer 6 108.8 Singer 7 191.3 Mean SD

33 13.7 40.1

262.4 293.2 393.8

147.4 41.5

112.5 82.4 94

316.5 68.7

9.5

55.3 51.7 62.5

17.2 19.8 12.4 16.0

Abbreviation: SD, standard deviation. The columns marked Re F0Speech (relative to F0 in speech) list the interval in semitones between the mean F0 in speech and in the Recitative and Song, respectively. MVF0, mean of F0; SDF0, SD of F0

increased to about 4 or 5 Hz toward the end of the tones. This case is illustrated by the case of the Old Man role sample in the figure. In this case, the mean peak-to-peak extent was about 3.7 semitones. In the Colorful Face role sample in the same figure, the corresponding value was much narrower, only 1.2 semitones. In this case, there was a slight upward drift at the end of the note, produced during the final consonant /n/ of the syllable, however without giving the impression of a new tone. The spectrum of some sustained vibrato-free tones showed an unbroken series of harmonic partials extending up to very

Speech

Recitative

0,04

Occurrence [%]

Occurrence [%]

0,05

high frequencies, sometimes as high as 17 kHz, as illustrated in Figure 5. These examples were taken from singer 6 and 8, who both performed the Colorful Face role. Figure 6 shows LTAS for Speech and Recitative for all singers, and the Song for the singers performing the Old Man and Colorful Face roles, respectively. All LTAS curves show a main peak below 1000 Hz. For Speech it is near 500 Hz and for Song it is near 800 Hz. Most curves for the Recitative are similar to those for the Song. However, unlike the LTAS curves for the Recitative and Song as performed by the Old Man role

0,03 0,02 0,01 0,00

-18

-12

-6

0

6 -12

6

12

18

Song, Colorful Face role

Song, Old Man role

0,04

0

Occurrence [%]

Occurrence [%]

0,05

-6

0,03 0,02 0,01 0,00

0

6

12

F0 [semitones re 220 Hz]

18 0

6

12

F0 [semitones re 220 Hz]

FIGURE 2. Typical examples of F0 histograms for the indicated cases.

18

140

Journal of Voice, Vol. 26, No. 2, 2012

singers. This difference would reflect a combination of a stronger voice source fundamental and lower formant frequencies in the Western baritone singer. Figure 8 illustrates how the spectrum slope varied with vocal loudness as measured in terms of Leq. The slope was measured as the trendline slope in the frequency range 700–6000 Hz. The low Leq values for Speech were associated with less steep slope values, whereas those for the Recitative and Song were associated with steeper slope values. There were no systematic differences between the two roles. The correlation between the slope and Leq is quite weak.

TABLE 3. Intervals in Semitones Between the Peaks in the F0 Histograms for the Song Old Man Role Singer Peak 1 2 3 4 5 6

2

3

0.00 2.72 4.98 6.58

3

0.00 0.00 1.93 3.41 3.97 5.50 6.84 7.12 8.73 10.38 12.0

Colorful Face Role 7

4

6

8

0 3.59 5.66 7.61

0 2.93 4.92 6.86

0 1.68 4.89 6.85 8.66

0 1.96 3.97

Ratios close to intervals commonly used also in Western music are given in bold.

singers, those of the Colorful Face role singers showed a blunt peak centered at 3 kHz, approximately, both in Recitative and Song. A similar difference can be noted also in the case of these respective singers’ Speech. For both roles, the average slope above the main peak was about 10 and 11 dB/octave for the Recitative and for the Song, respectively. Figure 7 shows the averaged LTAS curves for the two roles illustrating the difference in the frequency region near 3000 Hz. For comparison, the same figure also shows an LTAS of a professional Western opera baritone, the LTAS curve of whom shows a prominent peak near 2700 Hz surrounded by deep valleys. This peak, that is, the singers’ formant cluster, being typical of Western classically trained male opera singers, is much higher and narrower than any of those produced by the Peking opera singers (Figure 6). In the low frequency range and up to about 500 Hz, it can also be noted that the LTAS curve of this baritone is around 15 dB higher than those of the Peking

DISCUSSION Not unexpectedly, our results showed a number of striking differences between the classical Peking opera and Western opera singing. For instance, the Peking opera singers’ pitch range reached higher, the highest scale tone observed at 611 Hz, pitch about Eb5. The Western operatic tenors do not generally need to sing tones higher than C5.3 The songs analyzed were representative for the respective roles. Hence, analysis of the intervals between the scale tones seemed relevant. The lowest scale tones in the songs varied between 155 (pitch close to Eb3) and 378 Hz (pitch close to F#4). The mean interval between the lowest and highest F0 histogram peaks was no more than about seven semitones, that is, a fifth, and the maximum interval between scale tone peaks was 12 semitones, that is, just one octave. Western operatic singers are generally required to have a useful range of at least two octaves. It should be remembered, however, that these values represent nothing but the intervals between the lowest and highest clear peaks in the F0 histograms. The singers often sang pitches higher than the frequency of the highest scale tone. Also the average F0 in Speech was quite high, near 200 Hz, which clearly is much higher than what is typically found in Western male adults. Mean F0 in Chinese males’ spontaneous

14 12

Occurrence [%]

10 8 6 4 2 0

0

0,2 0,4 0,6 0,8

1

1,2 1,4 1,6 1,8 2

2,2 2,4 2,6 2,8

3

3,2 3,4 3,6 3,8

4

4,2 4,4 4,6 4,8 5

5,2 5,4 5,6 5,8

6

6,2 6,4 6,6 6,8 7

7,2

Interval [semitones]

FIGURE 3. Histogram of intervals, in semitones, observed within songs between all pairs of F0 histogram peaks.

Johan Sundberg, et al

141

Acoustical Study of Classical Peking Opera Singing

Level [dB]

-20

Singer 5, Colorful face

-40 -60 -80 -100 -120

0

5000

10000

15000

20000

Level [dB]

-20 Singer 8, Colorful face

-40 -60 -80 -100

FIGURE 4. Examples of vibrato in the songs as sung by Colorful

-120

Face role singer 4, and by Old Man role singer 1 (upper and lower panel, respectively).

0

5000

10000

15000

20000

Frequency [Hz]

FIGURE 5. Examples of spectra of loud tones sung by the indicated speech has been reported to be 126 Hz, SD 34.5 Hz.4 The elevated speech F0 used by these singers may belong to the voice characteristics required for a successful career as a performer in the classical Peking opera. According to Wang,1,5 classical Peking opera singers possess a ‘‘singer’s high formant.’’ As illustrated in Figure 7, our LTAS analyses showed no sign of this, neither for the Old Man role singers nor for the Colorful Face role singers. A marked spectrum peak was observed in the high frequency range only in the case of the song performed by the Colorful Face role singers, but its center frequency was located around 3300 Hz and was much lower in amplitude than the singer’s formant cluster seen in male classical Western opera singers. The discrepancy may have several reasons, for example, that Wang’s instead of LTAS analyzed LPC spectra, which represent an interpretation of the spectrum rather than the spectrum itself. For instance, it showed a spectrum envelope of an /i/ vowel where the F1 peak was 30 dB weaker than the F2 peak.5(p306) Thus, he applied a quite special definition of his term ‘‘singer’s high

singers. The noise floor of the recording was between 100 and 125 dB in the frequency range 16–20 kHz.

formant.’’ It is also possible that his subjects represented another singer tradition than the one analyzed here. In any event, our results indicate that the singers we analyzed did not sing with a singer’s formant cluster. The singer’s formant cluster has been shown to help the singer’s voice cut through the sound of a loud orchestral accompaniment, the point being that the peak appears in a frequency region where the sound of the orchestra is much softer than in the low frequency range, where the loudest sound levels occur. This holds for the Western classical music orchestra. The absence of a singer’s formant cluster in the Peking opera singers would be related to the enormous timbral differences between the orchestral accompaniments used in the Peking and Western opera traditions. These are illustrated in Figure 9, which compares LTAS data of Western symphonic orchestras and an example of Peking opera accompaniment. While for the

Speech

Recitativ 100 Singer 1 Singer 2 Singer 3 Singer 4 Singer 5 Singer 6 Singer 7

90

Mean level [dB]

Mean level [dB]

100

80 70 60 50

80 Singer 1 Singer 2 Singer 3 Singer 6 Singer 4 Singer 5 Singer 7

70 60 50

40 0

100

12

24

36

48

60

72

Old Man role, Song Frequency [semitones re 100 Hz]

84

40 96 0

12

24

36

48

60

Colorful Face role Song

72

84

96

Frequency [semitones re 100 Hz]

100

90

90

80 70 60

Singer 1 Singer 2

50

Singer 3 Singer 6

Mean level [dB]

Mean level [dB]

90

80 70 60

Singer 4 Singer 5

50

Singer 7

40

40

0 100

12 24 36 1600 48 3200 60 6400 72 12800 84 200 400 800 Frequency [semitones re 100 Hz] Frequency [Hz]

0 96100

12

200

24

36

48

60

72

84

400 800 1600 3200 6400 12800 Frequency [semitones Frequency [Hz] re 100 Hz]

FIGURE 6. LTAS of the indicated singers and tasks.

96

142

Mean spectrum level [10 dB/division]

Mean level [10 dB/division]

Journal of Voice, Vol. 26, No. 2, 2012

Old man, Singing Colourful face, Singing Western opera baritone

•

0

12

200

24

400

36

800 48 1600 60 3200

72 6400

84 12800 96

Frequency [Hz]

100

1000

10000

Frequency [kHz]

FIGURE 7. LTAS curves, averaged across Old Man role singers and Colorful Face role singers for the Song. The thin curve pertains to a professional Western opera singer.

FIGURE 9. LTAS of Western symphonic orchestras and of an exam-

Western opera, the spectrum level drops by about 11 dB/octave above 700 Hz, the spectrum for the Peking opera rises by about 6 dB/octave between 400 and 2000 Hz, and then drops by around 30 dB/octave above 2000 Hz. There would be little gain for a Peking opera singer to develop a Western singer’s formant cluster. The LTAS peak near 3300 Hz found in the Colorful Face role singers is similar to the peak that Leino and associates have found in ‘‘good’’ speaker voices6,7. A similar peak has been observed also in Western singers performing in the popular music genres.8,9 The origin of this peak is unclear. In the Colorful Face role singers, it was followed by a deep minimum in the LTAS curve, suggesting that the peak is produced by a clustering of the fourth and fifth formants. A remarkable finding is the presence of an unbroken series of harmonic partials reaching up to 17 000 Hz. In Western operatic singing, there is typically a marked minimum in the LTAS curve near 4000 Hz and above this frequency only groups of harmonic partials occur.10 This minimum is probably produced by the piriform sinuses.2,11 The Peking opera singers did show a minimum which however was less marked and which did not eliminate harmonic partials completely, as was shown in Figure 5. It occurred around 5000 Hz, thus suggesting that the piriform sinuses are smaller in these singers, presumably because of a higher larynx position and/or shorter and narrower vocal tracts.

Another remarkable finding was that the mean spectrum slope was steeper in the loud Recitative and Song samples than in the much softer Speech samples. Previous analyses have indicated that the gain in the higher overtones is greater than the gain in the lower overtones when vocal loudness is increased.12 In other words, the spectrum slope is normally steeper for softer than for louder voice. The reason for this is that the transglottal airflow is arrested less abruptly when subglottal pressure is decreased. The reason why the loud singing samples showed a steeper LTAS slope than the softer speech samples may be the substantial difference in mean F0 and perhaps also in formant frequencies between their singing and speech.

FIGURE 8. Relationship between spectrum slope and Leq for Speech (gray squares), and for Recitative and Song (open and filled symbols) performed by the Old Man role and by the Colorful Face role singers (circles and triangles).

ple of the instrumental accompaniment in a Peking opera (dashed and solid curves).

CONCLUSIONS In this study, some acoustic characteristics of classical Peking opera singing were analyzed and compared with those found in Western operatic singing. Peking opera singing was found to be around 15–20 dB louder than in the same singers’ neutral speech, and mean SPL at 0.3 m in singing was close to 95 dB. Pitch was almost one octave above that of speech in Old Man role singers and more than one octave in Colorful Face role singers. The F0 histogram for the songs revealed marked peaks separated by intervals, some of which agree with those found in Western music cultures: major second, pure fourth, and pure fifth. Vibrato rate was about 3.5 Hz, that is, 2 Hz lower than in Western classical singing. Often, it was speeded up toward the end of the tones. The spectrum of some vibrato-free tones contained an unbroken series of harmonic partials reaching up to 17 000 Hz, which is much higher than what is normally seen in Western operatic singing. The LTAS curves showed no trace of a singer’s formant cluster. The Colorful Face role singers’ LTAS showed a marked peak near 3300 Hz, somewhat similar to that found in Western pop music singers. The mean LTAS curve slope in range 700–6000 Hz was about 11 dB/octave in singing and tended to increase with increasing loudness. Thus, the spectrum slope was steeper in singing than in speech, presumably because F0 was much higher in singing. Acknowledgments The authors are indebted to Professor Tang, Yincheng, Department of Peking Opera, Chinese Opera Institute, Beijing, China

Johan Sundberg, et al

Acoustical Study of Classical Peking Opera Singing

for organizing the participation of the singers, and the Education Bureau of Beijing, China for economical support. The authors acknowledge the kind cooperation of the singers as well as the helpful comments and suggestions by two anonymous reviewers.

REFERENCES 1. Wang S. Singing voice: bright timbre, singer’s formants and larynx positions. In: Askenfelt A, Felicetti S, Jansson E, Sundberg J, eds. SMAC 83. Proceedings of the Stockholm International Music Acoustics Conference. Vol 1. Stockholm, Sweden: Royal Swedish Academy of Music: No. 46:1; 1984:313–322. 2. Sundberg J. Articulatory interpretation of the ‘‘singing formant’’. J Acoust Soc Am. 1974;55:838–844. 3. Titze IR. Principles of Voice Production. Englewood Cliffs: Prentice Hall; 1994. 4. Wang M, Lin M. An Analysis of Pitch in Chinese Spontaneous Speech. In: Archives of International Symposium on Tonal Aspects of Languages: With Emphasis on Tone Languages. Peking, China, March 2004:28–31.

143

5. Wang S. Singer’s high formant associated with different larynx position in styles of singing. J Acoust Soc Jpn (E). 1986;7:303–314. 6. Leino T. Long-term average spectrum in screening of voice quality in speech: untrained male university students. J Voice. 2009;23: 671–676. 7. Leino T, Laukkanen A-M, Radolf V. Formation of the actor’s/speaker’s formant: a study applying spectrum analysis and computer modeling. J Voice 2010; [Epub ahead of print]. 8. Zangger Borch D, Sundberg J. Some phonatory and resonatory characteristics of the Rock, Pop, Soul and Swedish Dance Band styles of singing. J Voice. 2010;25:150–158. 9. Cleveland TC, Sundberg J, Stone RE. Long-term-average spectrum characteristics of country singers during speaking and singing. J Voice. 2001;15: 54–60. 10. Ternstr€om S. Hi-Fi voice: observations on the distribution of energy in the singing voice spectrum above 5 kHz. In: Proceedings of Acoustics ’08. Paris: France; 2008:3171–3176. 11. Dang J, Honda K. Acoustic characteristics of the piriform fossa in models and humans. J Acoust Soc Am. 1997;101:456–465. 12. Sundberg J, Nordenberg M. Effects of vocal loudness variation on spectrum balance as reflected by the alpha measure of long-term-average spectra of speech. J Acoust Soc Am. 2006;120:453–457.