STUDY ON SUBJECTIVE PREFERENCE VALUES OF ACOUSTIC PARAMETERS FOR PEKING OPERA Lu Yang, Hui Ma, Jiaojiao Yang and Boya Yu Tianjin University, School of Architecture, 300072 Tianjin, China email: [email protected] Compared with the study on acoustic quality of western music and opera, few studies were carried out on how the contemporary people evaluate acoustic attributes of Chinese traditional opera. In this study the subjective preference values of acoustic parameters for Peking opera such as sound pressure level, reverberation time and frequency response curve of RT were revealed through laboratory experiments after a preliminary investigation. Both ordinary group and fans group participated in the experiments and the different rhyme scheme of Peking opera was considered. The results showed that the preferred sound pressure level of Peking Opera was from 70dBA to 73dBA, and the preferred RT value was from 0.2s to 0.6s. As for the frequency response curves of RT, more subjects preferred to an ascendant curve, which means longer reverberation time for high frequency components rather than low frequency ones. The results were quite different from the Chinese criteria on acoustic design for theatres.

1.

Introduction

There has been a long tradition of acoustic design for theatrical building in the world, and nowadays the amount of opera audience increases significantly and rapidly. Hence, developing a more comfortable acoustic environment for theatrical buildings is very important in modern society. Theatrical building design requires a strongly specific consideration of performance types, which means different types of opera require different levels of acoustic parameters. Opera has a long and great tradition in China, and Peking opera is a typical kind of traditional Chinese operas, whose performance combines music, vocal performance, mime, dance, and acrobatics altogether. It arose in the late 18th century and became fully developed and recognized by the mid-19th century. It has also spread to other countries such as the United States and Japan. Thus, it is a suitable objective to be researched for revealing acoustic requirements for opera houses employed for Chinese traditional operas. Since Peking opera has some distinctive characteristics, acoustic requirements for its performance places are different from them of concert halls, opera houses, theatres and so on [1]. As for studies on subjective evaluation to acoustic parameters, numerous studies were conducted on audiences’ evaluation of acoustical quality of different types of music in western countries. For example, Kuhl, et al. studied the preferred reverberation time and found that it was related with the types of music: for classical music, the result was 1.54s, the preferred RT for romance music was 1,54s and for modern music it was 1.84s [4]. Similar studies of Schroeder, et al. indicated that the advisable RT for classical music was 2s, while Ando, et al.’s research concluded that the preferable choice of RT for classical music was 1.2s [5, 6]. In China, similar studies on acoustical assessment about Chinese national music have been carried out as well [7, 8], but there have been no research focusing precisely on the evaluation of acoustic quality of traditional Chinese opera. In terms of places of opera performance, there have been some previous studies focusing on traditional Chinese theatre [2, 3]. However, studies aiming precisely on modern performance places of Peking opera are absent. 1

The 23rd International Congress on Sound and Vibration

The aim of this research was to find out the relationship between acoustic parameters and subjective evaluation of Peking Opera, which could be used to improve the acoustic quality of Peking opera theatre.

2.

Methodology

In this paper, laboratory studies were conducted to explore the subjective preference values of some key physical parameters selected by pre-investigation. According to a previous questionnaire survey, 3 major physical parameters were identified, which were sound pressure level, reverberation time and frequency response curve of RT. Therefore, there were three experiments focusing respectively on these three parameters in this whole study. With regard to each parameter, a subjective preferred experiment was carried out to find out its subjective preferred levels, values or types for Peking opera performance. 2.1 Optimization experiment 1：Sound pressure level In order to find out subjective preferred sound pressure levels for Peking opera, in these study subjects were asked to select preferable sounds from a bunch of experiment sounds with varying sound pressure levels of single experiment segments, by the paired-comparison method. When choosing original sound materials, in order to make these sound segments more representative, main features of Peking opera had been taken into account, which were roles of performers, stage speech or song, and musical speeds. Given the combination of these three features, the selected group of experiment segments was listed in Table 1. Dry signals of these experiment segments were recorded in a semi-anechoic chamber, where the background noise level was approximately 20dBA, which was low enough to make sure that the recording process was unaffected. This recording work was done by live performance, and by several professional Peking opera actors. As was shown in Table 2, Sound pressure levels applied in the formal experiment were determined according to Ando’s previous study on preferred sound pressure level for concert hall and pre-experiments[9], 4 levels had been chosen, those could represent the sound pressure level range of common Peking opera performance and be recognized the differences among them by subjects. In order to make sure that sound pressure level was the only parameter varied of experiment sounds to limit the analysis to this single parameter in this experiment, an acoustic software was applied to enable that sound pressure of experiment sounds was varied at 4 levels (63dBA, 70dBA, 73dBA and 76dBA) while the other parameters were fixed simultaneously through convolution calculation between dry signals and specific impulse responses. The other two studied parameters were set depending on the current Chinese criteria for theatre design: the RT value was 1s, and the type of frequency response curves of RT was “flat”. Given that different sound segments with different features of Peking opera performance might have different subjective preferred levels of sound pressure, all experiment sounds made by the same sound segment composed one session, and each session corresponded to one relatively independent experiment. There were 10 sessions out of the amount of experiment segments shown in table 1, and there were 4 different experiment sounds in every session out of the amount of sound pressure levels. In terms of subjects, for the sake of recognizing whether or not there were any differences between two different groups of subjects: the normal group including listeners with little experience of live performance of Peking opera and the fans group consisting of audience with abundant experience. Totally, there were 40 subjects participating in these experiments and the amount in each group was the same. All of them with various ages had normal hearing abilities. In the process of the subjective experiment, subjects were asked to wear monitoring headphones, sit in a semi-anechoic chamber, and listen experiment sounds with full concentration in a comfortable thermal environment. After hearing, they were required to evaluate these sounds. In accordance with the paired-comparison method, which is considered to be the most effective method for examining 2

ICSV23, Athens (Greece), 10-14 July 2016

The 23rd International Congress on Sound and Vibration

subjective preferences due to the fact that subjects are not required to make absolute judgments when using this method, each subject was supposed to compare 6 pairs per session for the reason that the amount of sound pressure levels was 4 and there were altogether 6 patterns when picking 2 different items from 4. A total of 10 sessions was conducted for each subject and each session corresponded to a single experiment segment. The interval between the experiment sound presentation was 10s, and that between comparison pairs was 30s to provide sufficient time for subjects to evaluate these sounds by ticking off an item between“sound1” and “sound 2”. The score of every sound pressure level in every session was obtained here by accumulating score giving +1 and -1 corresponding to positive and negative judgments, respectively. Table 1: The selected group of sound segments.

Roles Qingyi

Song

Xiaosheng Laodan Laodan Stage Speech

Yunbai Jingbai

No. A1 A2 A3 A4 A5 A6 A7 A8 A9 A10

Speed of Rhythms Slow Fast Slow Fast Slow Fast Slow Fast Slow Fast

Table 2: Considered sound pressure levels.

Experiments Ando Pre-experiments Formal experiments

Division modes of Sound pressure levels（dBA） 74 77 80 83 70 73 76 79 67 70 73 76

2.2 Optimization experiment 2：reverberation time (RT) Subjective preferred experiment on RT was similar with that on sound pressure level. Most part, including the selected group of sound segments, the experiment sound processing, the subjects and the subjective evaluation method of this experiment were the same as those on sound pressure level. The slight difference between these two experiments was as follows: Values of RT applied in this experiment according to previous studies and pre-experiments were listed in table 3[7, 8, 10]. Table 3: Considered values of RT.

Studies J. Sarwono

0

0.45

0.6

Considered values of RT (s) 1.2

2.5

4.5

Wu 0 0.4 0.8 1.2 1.8 2.5 4.5 Meng Lute 0.6 1.0 1.2 1.4 1.6 1.8 2 2.4 2.6 3 Flute Erhu 0.6 1.0 1.4 1.6 1.8 2 2.4 2.8 3.2 3.9 This study 0.2 0.6 1.0 1.4 1.8 2.5 The sound pressure level set in this experiment was 70dBA, which was the best sound pressure level according to the result of experiment 1 in this study. As for the frequency response curves of RT, it was the “flat” for the reason that it was the most common situation and this study had not found the preferred type when carrying out this experiment. ICSV23, Athens (Greece), 10-14 July 2016

3

The 23rd International Congress on Sound and Vibration

In addition, each subject in this experiment was supposed to compare 15 pairs per session for the reason that the amount of RT values was 6. 2.3 Optimization experiment 3：Frequency response curve of RT Subjective preferred experiment on frequency response curve of RT was similar with that on sound pressure level. Most part, including the selected group of sound segments, the experiment sound processing, the subjects and the subjective evaluation method of this experiment were the same as those on sound pressure level. The slight differences between these two experiments were as follows: Since relative studies on frequency response curve of RT was absent, this study set 5 representative types of frequency response curves of RT after considering all kinds of situations: as was shown in table 4, all their RT values over the medium frequency range(500Hz and 1000Hz) were 0.6s, which was the best RT from the result of the study on RT in this paper, and when it came into low frequency range(125Hz and 250Hz) and high frequency range(2000Hz and 4000Hz), these two values of the “Ascent” curve were respectively 0.4s and 0.8s so that the whole trend was increasing; as for “Descent” curve, they were 0.8s and 0.4s; in terms of the “Flat” curve, “U” curve and the “upside-down ‘U’ ” curve, they were 0.6s and 0.6s, 0.8s and 0.8s , and 0.4s and 0.4s, respectively. The subjective experiment processes of RT and frequency response curve of RT were the same as of sound pressure level. Table 4: Types of frequency response curves of RT.

Types of frequency response curves Ascent Descent Flat “U’’ Upside-down“U”

125 0.4 0.8 0.6 0.4 0.8

250 0.4 0.8 0.6 0.4 0.8

Frequency（Hz） 500 1000 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6

2000 0.8 0.4 0.6 0.4 0.8

4000 0.8 0.4 0.6 0.4 0.8

The sound pressure level set in this experiment was 70dBA, which was the best sound pressure level according to the result of experiment 1 in this study. As for RT, it was 0.6s for the same reason. In addition, each subject in this experiment was supposed to compare 10 pairs per session for the reason that the amount of types of frequency response curve of RT was 5.

3. Results and Discussions 3.1 Preferred sound level Firstly, in order to examine whether or not some considered factors could affect subjective evaluation results, variance analysis was carried out. In terms of sound pressure levels, these considered factors included sound pressure levels, genders of performers and groups of subjects. The result listed in table 5 indicated that sound pressure levels had a significant influence on subjective preference on sound pressure levels at a 0.05 significance level, and there was an interaction between genders of performers and sound pressure levels, which means that they had a combine effect on subjective preference on sound pressure levels. Based on the above results that there were differences of subjective preference evaluation depending on different performer genders, analysis aiming at finding preferred sound pressure levels was supposed to be carried out separately for each gender. By calculating the mean value, the score of every sound pressure level varying genders of performer was obtained, which was shown in Figure 1. All sounds with the highest mean value varying the performer genders were the sounds with a sound pressure level of 70dBA. And by means of respective T-text analysis, whether or not these sounds with a sound pressure level of 70dBA had significant differences with those with the other levels of sound pressure was investigated. The levels of sounds with the highest mean value and

4

ICSV23, Athens (Greece), 10-14 July 2016

The 23rd International Congress on Sound and Vibration

sounds with no significant differences with them were the subjective preferred sound pressure levels, which were 70dBA for female, and 70dBA and 73dBA levels for male. Table 5: Effects of considered factors on preferred sound pressure levels.

Source

Significance

Sound pressure Sound pressure * Groups of subjects Sound pressure * Genders Sound pressure * Groups of subjects * Genders

.000 .270 .007 .845

Mean values of subjective evaluation score for different SPLs

5 4

male

3

female

2 1 0 -1 -2 -3

-4 -5 67

70

73

76

SPL(dBA)

Figure 1: Mean values of subjective evaluation score for different sound pressure levels.

3.2 Preferred reverberation time (RT) The result of variance analysis on RT indicated that RT values had a significant influence on the preferences of subjects at the 0.05 significance level, meanwhile roles, speeds of rhythms and groups of subjects had an interaction with sound pressure levels and they had a combine effect on the preference values. Based on the above results, analysis aiming at obtaining subjective preferred RT values varying roles, speeds of rhythms and groups of subjects was carried out separately. By calculating the mean value, the score of every RT value varying roles, speeds of rhythms and groups of subjects was obtained, which was reported in Figure 2. Most of sounds varying roles, speeds of rhythms and groups of subjects with the highest mean value were the sounds with a RT value of 0.2s, while the highest mean RT value of sounds with a slow speed and a “sheng” role was 0.6s. And in accordance with respective T-text method, whether or not these sounds with a RT value of 0.2s or 0.6s had significant differences with those with the other RT values was investigated. As was shown in table 6, RT values of sounds with the highest mean value and sounds with no significant differences with them were the subjective preferred RT values, which were 0.2s, 0.6s or 0.2s and 0.6s for most items. Preferred RT values of sounds with a “Laosheng” or “Laodan” role and a relatively slow musical speed might be 1.0s, 1.4s or 1.8s as well.

ICSV23, Athens (Greece), 10-14 July 2016

5

Mean values of subjective evaluation score for different RT values

The 23rd International Congress on Sound and Vibration

5

qingyi-fast

qingyi-slow

4

xiaosheng-fast

xiaosheng-slow

laosheng-fast

laosheng-slow

laodan-fast

laodan-slow

stage speech-yunbai

stage speech-jingbai

3 2 1 0

-1 -2 -3 -4 0.2

0.6

1

RT(s)

1.4

1.8

2.5

Figure 2: Mean values of subjective evaluation score for different RT values. Table 6: Subjective preference values of RT.

Experiment Samples Qingyi-Fast Qingyi-Slow Xiaosheng-Fast Xiaosheng-Slow Laosheng-Fast Laosheng-Slow Laodan-Fast Laodan-Slow

Stage speech-Yunbai Stage speech-Jingbai

6

Groups of subjects Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal Fans Normal

Preferred Values 0.2s、0.6s 0.6s 0.2s、0.6s 0.6s 0.2s 0.6s 0.2s、0.6s 0.2s、0.6s 0.2s、0.6s 0.2s、0.6s 0.2s、0.6s、1.0s 0.6s 0.2s、0.6s 0.2s、0.6s 0.2s、0.6s、1.0s、 1.4s、1.8s 0.6s、1.0s、1.4s 0.2s、0.6s 0.2s、0.6s 0.2s 0.2s、0.6s

ICSV23, Athens (Greece), 10-14 July 2016

The 23rd International Congress on Sound and Vibration

Mean values of subjective evaluation score for different frequency response curves of RT

3.3 Preferred frequency response curves of RT The result of variance analysis on frequency response curves of RT indicated that different types of frequency response curves of RT had a significant influence on the preferences of subjects at the 0.05 significance level, meanwhile roles and speeds of rhythms had an interaction with sound pressure levels, and they had a combine effect on the preference values. Based on the above results, analysis aiming at obtaining subjective preferred types of frequency response curves of RT varying roles and speeds of rhythms was carried out separately. By calculating the mean value, the score of every type of frequency response curves of RT varying roles and speeds of rhythms was obtained, which was shown in Figure 3. Most of sounds varying roles and speeds of rhythms with the highest mean value were the sounds with a “ascend” or a “descent” frequency response curves of RT. And in accordance with respective T-text method, whether or not these sounds with a “ascend” or a “descent” frequency response curves of RT had significant differences with sounds with the other types of frequency response curves of RT was investigated. As was shown in table 7, frequency response curves of RT of sounds with the highest mean value and sounds with no significant difference with them were the preferred types of frequency response curves of RT, which were “ascend” or “descent” in most cases, in addition, preferable types of frequency response curves of RT of sounds with a “Laosheng” role or the stage speech part might have some other options. qingyi-fast xiaosheng-fast laosheng-fast laodan-fast

1.5 1

qingyi-slow xiaosheng-slow laosheng-slow laodan-slow

0.5 0 -0.5 -1 -1.5 -2 ascent

descent

“U”

upside-down “U”

Flat

Frequency response curves of RT

Figure 3: Mean values of subjective evaluation score for different types of frequency response curves of RT. Table 7: Subjective preference values of frequency response curve of RT.

Experiment Samples Qingyi-Fast Qingyi-Slow Xiaosheng-Fast Xiaosheng-Slow Laodan-Fast Laodan-Slow Laodan-Fast Laodan-Slow Stage speech-Yun Stage speech-Jing

ICSV23, Athens (Greece), 10-14 July 2016

Preferred Values Ascent、Descent Ascent、Descent、Flat “U’’、Upside-down “U”、Flat Ascent、“U’’、Flat Ascent、Descent Ascent、Descent Ascent、Descent Ascent、Descent Ascent、“U’’、Flat Ascent、Upside-down “U”

7

The 23rd International Congress on Sound and Vibration

4. Conclusions The results showed that the preferred sound pressure level of Peking Opera was from 70dBA to 73dBA, and the preferred RT value was from 0.2s to 0.6. 0.2s was better in most cases except the “sheng” role with a slow speed, in that case 0.6s won the highest score. In addition, there were some differences with the different groups of subjects: the fans group preferred a longer RT compared to the normal group. This result was quite different from the Chinese criteria on acoustic design for theatres (0.8s-1.2s) [11]. As for the frequency response curves of RT, more subjects preferred to an ascendant curve, which means longer reverberation time for high frequency components rather than low frequency ones. The descendent curve was also acceptable for most performance types. However, performance of “Xiaosheng” with a fast speed showed differences with this general conclusion, subject preferred the other three types under this condition.

5. Acknowledgment This work was supported by the National Natural Science Foundation of China (Project No: 51178300 and 51478303). The author gratefully acknowledged assistance of my colleagues, Peking opera actors and fans who participated in the study, without whom, this study would not been possibly conducted.

REFERENCES 1 Wang, J. Q. Acoustical design of building halls, Tianjin Science and Technology Press, Tianjin (2001). 2 Wang, J. Q. Acoustics of courtyard spaces, Acta Acustica, 4, 289–294, (2007). 3 Pu, H. J., Min, H. Q., Qiu, X. J., and Wang, J. Q. A study on the sound field in a vault with two open ends, Acta Acustica, 3, 30–50, (2009). 4 Kuhl, W. Über Versuche zur Ermittlung der günstigsten Nachhallzeit groβer Musikstudios, Acta Acustica, 4, 618-634, (1954). 5 Schroeder, M. R., Gottlob, D. and Siebrasse, K.F. Comparative study of European concert halls: correlation of subjective preference with geometric and acoustic parameters, Journal of the Acoustical Society of America, 56, 1195-1201, (1974). 6 Ando, Y., Okura, M. and Yuasa, K. On the Preferred Reverberation Time in Auditoriums, Acustica, 50, 134-141, (1982). 7 Wu, S. X. Investigation into Acoustics of Chinese National Music Halls, Journal of south China university of technology (natural science edition), 35, (2007). 8 Meng, Z. H. and Zhao, F. J. Preliminary test on the preferred reverberation time for the Chinese instrumental music motifs, Applied Acoustics, 1, 41-45, (2007). 9 Ando, Y. Concert hall acoustics, Berlin: Springer-Verlag, (1985). 10 Sarwono, J. and Lam, Y. W. The Preferred Acoustic Parameters for a Javanese Gamelan Performance Hall, Proc. 17th ICA, Rome, (2001). 11 Standard, GB/T50356. Code for architectural acoustical design of theater, cinema and multi-use auditorium, Standardization Administration of the People's Republic of China, (2005).

8

ICSV23, Athens (Greece), 10-14 July 2016