Temporal and spectral characteristics of public address announcements in subway platforms with various finishing materials using computer simulation

Transcription

1 Temporal and spectral characteristics of public address announcements in subway platforms with various finishing materials using computer simulation Yong Hee KIM a) Yoshiharu SOETA b) Health Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Midorigaoka, Ikeda, Osaka , Japan In subway platforms, the sound signals and announcements, containing urgent information such as an accident as well as indications of train-coming or destination, are delivered through public address (PA) loudspeakers as an important sound source. Various acoustic environments by platform designs, such as dimension or materials, are closely related to propagation and perception of PA announcements. This study investigates the effects of platform design elements such as finishing materials on temporal and spectral characteristics of PA announcements. In order to alter design condition systematically, two subway spaces with island platform are constructed using computer simulation. Then, it is acoustically matched to the field measurement results in terms of room acoustic parameters. Sound sources of the announcements are collected according to the announcing purposes (train movements, door operation and others). Auralized announcements in simulation model are evaluated in terms of sound pressure levels and auto-correlation/interaural cross-correlation factors in both time and frequency domains according to various combination of finishing materials (floor, wall and ceiling). As results, the acoustic design directions of platform spaces is discussed for improving sound quality and speech intelligibility of PA announcements. 1 INTRODUCTION Public address (PA) system for speech in train stations has an important role to provide proper travel information including indications of train-coming or train destination, and urgent instructions such as an accident 1,2. Especially, underground stations have worse acoustic environments, such as higher train noise level 3 and lower Speech Transmission Index (STI) with longer reverberation 4, than above-ground stations due to its long enclosure with reflective surfaces. Thus, improving speech intelligibility of PA system has been one of the important a) yh.kim@aist.go.jp b) y.soeta@aist.go.jp

2 issues for designing acoustic environments in subway stations 1,2. Since the theory of classical room acoustics is not applicable for long enclosures 5, the previous studies suggested the prediction methods of platform sound fields using numerical calculation 4,6, ray-tracing simulation 1,2,7, or scale model testing 8 for the evaluation of speech intelligibility of PA system. Currently, ray-tracing simulation has been used most frequently for investigation of station acoustics in relation to design process due to its easiness and practicality 7,9-11. However, considerations of the actual details, such as column or stair, and acoustic validation with the field measurements are still lacking in the previous studies. Therefore, accurate models for platforms should be developed based on the actual conditions. In order to improve sound fields in stations, architectural treatments can be good solution before the construction. Chew 9 investigated acoustic design approach of the integrated bus/rail station to decrease background noise levels by increasing ceiling height or reducing concourse width using numerical calculation. Haan 10 investigated several design elements such as noise barrier height or materials in high-speed train station using ray-tracing simulation. Especially, Kang 12 reported that the speech intelligibility of PA system could be improved by installation of diffusers and absorbers through scale model testing. However, the effects of architectural treatments including dimensional alterations or finishing materials are not fully investigated in subway platforms, in particular, with detailed models. The architectural treatments also affect propagation and perception of PA announcements as well as its speech intelligibility. Sato et al. 13 reported some temporal characteristics of PA announcements through field measurement. However, there is rare study on the temporal and spectral characteristics of spoken announcements over the PA system. In this study, therefore, architectural design elements for improving platform sound fields and acoustic characteristics of PA announcements are investigated in a ray-tracing simulation model. First, two simulation models of the subway platform are constructed based on the field measurement data to derive accurate simulation configurations. Then, effects of dimensional alteration and finishing materials on sound fields in platform are investigated in a station model. Finally, temporal and spectral characteristics of PA announcements are analyzed using simulated impulse responses in comparison with the acoustic properties of the collected PA sources. 2 SOUND FIELDS IN SUBWAY PLATFORMS 2.1 Target venues and 3D modeling Two subway stations in Japan were selected for the experiments. The field measurement results were described in the previous study 14. Station A as shown in Fig. 1 has two tracks and circular sectional tunnels, whereas station B as shown in Fig. 2 has three tracks and rectangular sectional tunnels, as a starting station. Both stations have island platforms, which has a single platform positioned between two parallel tracks. Based on the actual drawings, 3D models of stations A and B were built using computer-aided design software. In the 3D model, all elements of floor, wall, simplified ceiling were considered including elevator box, column, bulletin board, simplified stairways. Total length of platform is 162 m and 332 m for stations A and B, respectively. In the 3D model, modeled tunnel length is 112 m and 101 m for stations A and B, respectively. Cross-sectional dimensions of stations A and B are described in Table 1 and Fig Acoustic fitting In the field measurements, omni-directional loudspeaker was used as a sound source, which

3 representing PA loudspeaker at the longitudinal middle of platform. Height of the source was 2.2 m from the platform floor. A dummy head microphone was located at the standing passenger s ear level as a receiver. Four receiving positions were considered with 5 m spacing, and the distance between source and the first receiver was 5 m. In the simulation, 16 receiver points with 2.5 m spacing were considered including the field measurement positions. Fig. 4 shows the source and receiver positions in the simulation model. For acoustic fitting, commercial ray-tracing software (Odeon v.11.20) was used. This software has been widely used to predict acoustic environments of train station as well as auditoria 2,10,11. The details of the calculation parameters are listed in Table 2. From the simulation, binaural impulse responses were derived. Acoustic fitting was carried out by trial and error method in terms of five acoustic parameters specified in ISO : Early decay time (EDT), Reverberation time (RT, T30), Definition (D50), Clarity (C50) and Interaural crosscorrelation coefficient (1-IACCE3). Judgment whether the model is acoustically fitted or not is made based on the just noticeable difference (JND) values 15. Table 3 describes the results of acoustic fitting process in stations A and B. Difference of all five parameters between measurement and simulation was within JND values. The final acoustic input for each surface of the fitted model is derived in consideration of both station models A and B, as shown in Table Experimental configurations In this paper, two architectural treatments are considered for station model A: dimensional alteration and changes of finishing materials. For dimensional alteration, range of station dimensions for the current 12 underground stations in Japan was surveyed based on the previous study 3. Since room volume varied according to different dimensions, cross-section area (S) is also calculated for volumetric comparison. Station width (Ws) ranged 12.6 to 17.1 m (variation: 4.5 m), platform width (Wp) ranged 5.4 to 10.0 m (variation: 4.6 m), station height (Hs) ranged 4.3 to 5.9 m (variation: 1.6 m), corridor height (Hc) ranged 2.5 to 4.0 m (variation 1.5 m), and S ranged 50.3 to 69.7 m 2 (variation: 19.4 m 2 ). Thus, the dimension of station A is distributed within normal range as shown in Table 1. Based on the ranges of the current stations above, various combinations of station dimension are prepared as shown in Table 5. As for finishing materials, the current finishing materials in underground stations were surveyed 3,16. Table 6 describes the simulation configuration for alteration of finishing materials by six elements. Total absorption area (A) was calculated for each combination to investigate the effects of absorption on the sound fields by finishing elements. Absorption coefficients of the treated materials are referred to the previous data 17. For plotting the simulation results, a single value averaged from the all 16 receivers was used. C50 and IACC were averaged from 500 to 2k Hz. RT and EDT were averaged from 500 to 1k Hz. As for IACC, it was analyzed into two parts: early reflections (0 to 80ms; 1-IACCE3) and late reflections (80 ms to end; 1-IACCL3). The simulation situation is no passenger condition on the platform as is the field measurement condition. 2.4 Results From the simulation, the results of STI, C50 and D50 were very similar, and the results of RT and EDT were also very similar. Thus, the results of STI, EDT and 1-IACCE3 are plotted in Fig. 5 to 7. As shown in Fig. 5, STI tends to increase according to smaller cross-section area and larger absorption area. As shown in Fig. 5(a), smaller cross-section indicates reducing room volume so that reverberation decreases. From the linear regression analysis between STI and dimensional

4 factors, Ws and Hc were the most dominant factors for deriving STI within confidence level of 95%. However, platform width and station height showed no correlation with STI. As shown in Fig. 5(b), more absorption decreases reverberation so that clarity factors increase. From the linear regression analysis between STI and treated elements, all elements showed significant correlation with STI, except for track wall (W/T). In particular, absorption on the floor (F/P) and ceiling (C/P) in platform showed the dominant effects on STI. Before the absorption treatment, STI was fair level, but could be improved as good level with additional absorption of 300 m 2, roughly. As for, the reason for the insignificant of the treatments on the track wall (W/T) seems that the current corrugated surfaces or highly absorptive materials did not reflect sounds to platform horizontally. Comparing Fig. 5(a) and 5(b), treatments of finishing materials were much effective to improve STI. Fig. 6 shows that EDT follows reverse tendency with STI. From the linear regression analysis, Ws, Hc, F/P and C/P were found as the most dominant factors similarly to the STI results. As for the EDT results, the cases of station width and height also showed significant correlation with EDT. Since EDT almost follows the results of STI, the acoustic design of platform can be carried out based on STI. As shown in Fig. 7, both larger cross-section area and more absorption area commonly caused decrease of binaural dissimilarity in early reflections (1-IACCE3). Station width related dimensions (Ws and Ws+Hs) and all elements in platform (W/P, F/P and C/P) showed significant correlation with 1-IACCE3. As for binaural dissimilarity in late reflections, floor and ceiling in both platform and track (F/P, F/T, C/P, C/T) showed significant correlation with 1-IACCL3, but all station dimensions shows no correlation with 1-IACCL3. 3 PUBLIC ADDRESS ANNOUNCEMENTS IN SUBWAY STATION 3.1 Analysis methods Spectrogram of sound pressure level (SPL) using the discrete Fourier transform (FFT size=4,096 samples (=85.3ms), bandwidth=21.2 Hz, Hamming window, time grid spacing=10 ms) was generated for the time-frequency analysis. Additionally, auto-correlation/interaural cross-correlation factors (ACF/IACF) were calculated as a function of time to investigate sound quality of the PA sounds. In the short-time ACF/IACF calculation, selection of the proper integration interval (2T) for the target sounds is important to detect changes of sound quality characteristics 19. In this study, the integration interval of 30 ms and the running step of 10 ms were determined after the consideration of various time windows (15 to 1,920 ms) for the auralized PA sounds. From the ACF, τ 1 and φ 1 are derived to evaluate the perceived pitch and pitch strength, respectively, and IACC from the IACF is derived to evaluate the subjective diffuseness 3. The derived τ 1 and φ 1 from the both left and right channels were averaged as a single number. In order to investigate individual speech intelligibility of the selected impulse responses, speech-weighed C50 (wc50) is calculated with weighting factors of 15%, 25%, 35% and 25% for 0.5, 1, 2 and 4 khz, respectively 20. A-weighted sound pressure levels of the room impulse responses were also calculated. 3.2 PA messages Regular messages of the PA announcements in platforms are classified into the followings according to the announcing purpose: 1) notice of train schedule and acknowledgements (Ex. Thank you for using OO this time again), 2) notice of line/platform name, arrival/departure time, train type, destination (Ex. Shortly, a train bound for OO will be arrived at platform no.o), and 3)

5 other information including safety caution (Ex. Please step back from the yellow line for safety). According to the subway company s policy, they use male or female voice. The current PA messages were collected based on the personal recording of citizens in openair condition of the above-ground station 18. In open-air condition, signal-to-noise (S/N) ratio was more than 40 db to utilize the convolution with simulated room impulse responses. Fig. 8 describes the temporal and spectral characteristics of a sample PA announcement in open-air condition. Each sentence was concisely composed. According to the contents, there was an interval of 300 to 400 ms in-between two continuous spoken sentences. The most dominant frequency was 250 to 500 Hz for all PA announcements. 3.3 Experimental setup For the auralization of PA announcements, representative room impulse responses were selected from the simulation configuration specified in 2.3. Among 16 receivers, R5 was selected for the controlled comparison. Since distance between source and R5 was 15 m, this position is assumed to be sufficiently influenced by room conditions. From the alteration of station dimension, six cases with the largest dimensional variation by dimensional factors were selected. Another six cases with the large absorption variation by treated elements were selected. The selected 12 cases are asterisked in Tables 5 and 6. The reference sentence of 6 s with female voice was selected from the recorded sounds in open-air condition. The sentence contains the information of line, type and destination of upcoming train. In addition, frequency response of TOA CM-1801, which actually installed in station A, was measured to set-up the power level of sound source in ray-tracing simulation. Table 7 shows the measured sound power levels of the PA loudspeakers for deriving room impulse responses. 3.4 Results Fig. 9 shows the spectrograms of the left channel of the auralized PA announcements. The temporal and spectral characteristics of the auralized announcement, which convolved with the impulse response without any treatment (case of Current ), were blurred in comparison with the open-air condition. The intervals in speech were masked with the reverberation, so that S/N ratio is decreased by more than 20 db. Since the Current case has poor speech intelligibility (wc50 under -2 db), an improvements of the auralized announcements were expected through architectural treatments with better speech intelligibility. As results, the envelopments of the auralized announcements were significantly improved only by treatments of finishing materials except for W/T case. This is the same consequences from the evaluation results of sound field characteristics as specified in chapter 2.4 and Table 8. In particular, the cases of F/P, F/T, C/P and C/T showed clear temporal envelopments with S/N ratio of 30 db or more. However, spectral envelopes were not much changed by architectural treatments. From the dimensional treatments, SPL at both low- (125 to 250 Hz) and high-frequency (2 to 4 khz) bands was relatively decreased by around 1 db in comparison with SPL of the Current case. Contrarily, from the most treatments with absorptive materials, SPL at overall bands was decreased by around 3 db. Especially, decrease at mid-frequency bands (0.5 to 1 khz) was dominant. These tendencies are related to frequency characteristics of the treated materials. In brief, temporal modification was remarkable than spectral modification according to those architectural treatments since both room volume from dimension and absorption from materials are related to reverberation. In this study, experimental conditions related station dimension were designed to enlarge room volume, which leads to decrease STI and SPL as shown in Table 8, due to the

6 practical conditions such as minimum width for train. If station dimension is downscaled possibly, better speech intelligibility with reinforced early reflections is expected without absorption treatments. However, since smaller dimension might cause amplification of train noises, it is also needed to consider the effects of dimension on the background noises in the practical application. The results of ACF/IACF analysis for the auralized PA announcements are described in Table 8. As results, the cases related to station width and height (Ws and Hs) showed significant differences in τ 1 and φ 1 values in comparison with the Current case. Through these treatments, τ 1 increased by 2% to 3%, and φ 1 decreased by 3% to 6%. This indicates that larger crosssection of the station in terms of width and height can contribute to decrease the pitch of the PA sounds with smaller pitch strength. However, no significant changes were observed in IACC values by dimensional alterations. Contrarily, most architectural treatments by finishing materials, except for W/T case, showed significant differences in φ 1 and IACC values in comparison with the Current case. Through these treatments, φ 1 increased by 6% to 18%, and IACC increased by 14% to 25%. This indicates that absorptive materials on all materials in platform, and floor and ceiling in tunnel can contribute to increase the pitch strength of the PA sounds without any changes in pitch. In addition, IACC, which means lower binaural dissimilarity, can contribute to improve sound localization of the PA sources. Considering all treatments, φ 1 and IACC values of the auralized sounds showed significant correlations with STI and wc50 values of the room impulse responses (p<0.05). This reveals the possibility that speech intelligibility may be related to pitch strength and spatial impression. This can be verified through subjective investigations. 4 CONCLUDING REMARKS In this study, effects of station dimension and materials on sound fields and auralized PA announcements were investigated in order to provide acoustic directions of subway platform spaces. For this purpose, acoustic modeling data of station platforms for ray-tracing simulation were derived based on field measurements. Based on the experimental results from the computer simulation, strategy of acoustic design of subway platform for improving speech intelligibility of the PA system can be described as the followings: Minimize station dimensions with smaller room volume (Especially, station width and corridor height) to maximize STI, C50 and SPL of the PA system. Install absorptive materials first on floor and ceiling surfaces in platform to maximize STI, C50 of the PA system. The architectural treatments in subway platform had mainly influenced the temporal variation of the PA announcements by reverberation. Thus, time-related speech characteristics such as speech rate are important to prepare speech sources. Moreover, in real situation, S/N ratio of the PA system must be worse than the results of this study due to other noise sources. Therefore, more investigations with combined noise sources such as train car are required. On the other hands, changes of spectral envelopments according to architectural treatments are small, but frequency characteristics of the sound fields can be controlled by materials absorption properties. Consequently, it is needed to conduct subjective tests using the auralized PA sounds in order to find the desired sound quality characteristics of the PA system for passengers. 5 ACKNOWLEDGEMENTS This work was supported by a Grant-in-Aid for Young Scientists (A) from the Japan Society for the Promotion of Science ( ).

7 6 REFERENCES 1. P. A. A. Kootwijk, The speech intelligibility of the public address systems at 14 Dutch railway stations, J. Sound Vib., 193(1), , (1996). 2. Sander J. van Wijngaarden and Jan A. Verhave, Prediction of speech intelligibility for public address systems in traffic tunnels, Appl. Acoust., 67(4), , (2006). 3. Ryota Shimokura and Yoshiharu Soeta, Characteristics of train noise in above-ground and underground stations with side and island platforms, J. Sound Vib., 330(8), , (2011). 4. M. F. Harrison, Calculating speech intelligibility for the design of public address systems at railway stations, Proc. Instn. Mech. Engrs., Part F, 215, , (2001). 5. Jian Kang, The Unsuitability of the Classic Room Acoustical Theory in Long Enclosures, Arch. Sci. Rev., 39(2), 89-94, (1996). 6. Jian Kang, A method for predicting acoustic indices in long enclosures, Appl. Acoust., 51(2), , (1997). 7. Lening Yang and Bridget M. Shield, The prediction of speech intelligibility in underground stations of rectangular cross section, J. Acoust. Soc. Am., 109(1), , (2001). 8. Jian Kang, Scale modelling for improving the speech intelligibility from multiple loudspeakers in long enclosures by architectural acoustic treatments, Acta Acustica united with Acustica, 84(4), , (1998). 9. C. H. Chew, Integrated bus/rail station, Appl. Acoust., 56(1), 57-66, (1999). 10. C. H. Haan, Predicted effect of station design changes on high speed train noise, Building Acoust., 9(4), , (2002). 11. Zühre Sü and Mehmet Çalışkan, Acoustical Design and Noise Control in Metro Stations: Case Studies of the Ankara Metro System, Building Acoust., 14(3), , (2007). 12. Jian Kang, Improvement of the ST1 of Multiple Loudspeakers in Long Enclosures by Architectural Treatments, Appl. Acoust., 47(2), , (1996). 13. Takashi Sato, Hayato Sato, Hiroshi Sato and Masayuki Morimoto, Sound environment for speech communication at railway stations in Japan, Proc. Int. Congress on Acoust. (ICA2004), II, , (2004). 14. Ryota Shimokura and Yoshiharu Soeta, Evaluation of speech intelligibility of sound fields in underground stations, Acoust. Sci. & Tech., 32(2), 73-75, (2011). 15.ISO , Acoustics - Measurement of room acoustic parameters - Part 1: Performance spaces (2009). 16. Sang-Woo Lee, Dae-up Jeong, Su-Yaul Lee, Sang-Gon Cha and Myung-Suk Sym, A study on the evaluation of acoustic characteristic in underground subway stations in Seoul, J. Arch. Inst. Kr. in Korean, 19(4), , (2003). 17. M. David Egan, Architectural Acoustics, McGraw-Hill, United Kingdom, (1998).

8 K. Mouri, K. Akiyama and Y. Ando, Preliminary study on recommended time duration of source signals to be analyzed, in relation to its effective duration of the auto-correlation function, J. Sound Vib., 241(1), 87-95, (2001). 20. L. Gerald Marshall, Speech intelligibility prediction from calculated C50 values, J. Acoust. Soc. Am., 98(5), , (1995). Fig. 1 - Floor plan of 3D model for station A. Fig. 2 - Floor plan of 3D model for station B. Table 1 Cross-sectional dimensions of stations A and B. Ws and Wp indicate width of station and platform, respectively. Hs and Hc indicate height of station and corridor, respectively. S indicates cross-section area. Station name Ws [m] Wp [m] Hs [m] Hc [m] S [m 2 ] Station A Station B 14 to 17 5 to to 78.6 Fig. 3 - Cross-section of station A Fig. 4 - Source and receiver positions in simulation models. Red circle indicates sound source, and blue solid circles indicate receivers. Especially, blue dotted circle indicates the receiver positions in the field measurments

9 Table 2 The details of the calculation parameters for stations A and B. Station name Impulse response length [ms] Number of late rays Number of early rays Transition order Number of early scatter rays per image source Scattering methods Station A 5,000 28,663 57, Actual Station B 5,000 46,490 92, Actual Table 3 Acoustic fitting results for stations A and B. Averaged Station A Station B Acoustic JND frequency Measurement (M) ation (S) (S-M) ment (M) ation (S) (S-M) Simul- Difference Measure- Simul- Difference parameter values range EDT [s] Rel.5% (3%) (4.5%) RT [s] 0.5-1kHz Rel.5% (4.7%) (0.8%) D50 [%] C50 [db] 1 db kHz 1-IACCE Table 4 Acoustic input for each surface element of the fitted model. (Abbreviations- con c: concrete, w/: with, air: air cavity, h/w: hardware, st l: steel, Sc.: scattering coefficient) Elements Details Absorption coefficients by frequency 125Hz 250Hz 500Hz 1kHz 2kHz 4kHz Sc. Platform Wall Metal panel on con c wall Ceiling Suspended metal panel w/air Floor Granite tile w/ metal Bulletin board Plastic h/w on metal frame Elevator Glass pane w/ st l h/w w/air Column Absorptive finish Shutter Thin metal w/ large air Lower wall Patterned con c Tunnel Walls Corrugated rough con c w/ h/w Ceiling Con c surface with iron rail h/w Floor (Track) Iron rail on sleepers on con c Tunnel end Assumption of rare absorption Table 5 Simulation configuration for alteration of station dimension. (*: cases for auralization) Factor Ws [m] Wp [m] Hs [m] Hc [m] S [m 2 ] Ws 14.8/16.8/18.8/20.8* to 92.8* Ws+Hs 14.8/16.8/18.8/20.8* /5.6/5.9/6.2* to 104.6* Ws+Wp 14.8/16.8/18.8/20.8* 7.6/9.6/11.6/13.6* to 79.0* Wp /9.6/11.6/ to 92.8* Hs /5.6/5.9/6.2* to 67.4* Hp /3.3/3.6/3.9* 61.0 to 67.8*

10 Wall Floor Ceiling Table 6 Simulation configuration for alteration of finishing materials. /P and /T indicate platform and track area, respectively. (*: cases for auralization) Elements Reflective <<< Finishing materials (absorption area [m 2 ]) <<< Absorptive W/P Marble Ceramic tile Current (metal) Metal w/air Wood Wood w/air Drapery* (7 cases) (1,117) (1,119) (1,184) (1,186) (1,228) (1,404) (1,718) W/T Current (100% Con c) 50% Con c+50% Ballast 100% Ballast* (3 cases) (1,184) (1,658) (2,060) F/P Current (Granite) Rubber tile Carpet* (3 cases) (1,184) (1,237) (1,585) F/T Current (100% Con c) Con c block Mineral wool* (3 cases) (1,184) (1,357) (1,585) C/P Gypsum board Current (Metal sheet) 50% Acoustic board 100% Acoustic board* (4 cases) (1,162) (1,184) (1,550) (1,613) C/T Current (Con c) 50% Sprayed cellulose fiber 100% Sprayed cellulose fiber* (3 cases) (1,184) (1,497) (1,926) (a) Effects of station dimension Fig. 5 - Simulation results of STI. (b) Effects of finishing materials (a) Effects of station dimension Fig. 6 - Simulation results of EDT (b) Effects of finishing materials

11 (a) Effects of station dimension (b) Effects of finishing materials Fig. 7 - Simulation results of 1-IACCE3 (a) Acknowledgements (b) Notice of the coming train, destination (c) Safety caution Fig. 8 - Temporal and spectral characteristics of PA announcements in open-air condition Table 7 Sound power levels of the PA loudspeaker in simulation Frequency bands [Hz] k 2k 4k 8k Gain level [db] Table 8 STI and speech weighted C50 and A-weighted SPL of the impulse responses (R5 position) used for auralization, and ACF/IACF parameters of the auralized PA announcements. The asterisked cases showed significant differences from the Current case (t-test, p<0.05). Target signal Sound fields PA sounds Station dimension Finishing materials Factors Current Ws Ws (channel) Ws Wp Hs Hc W/P W/T F/P F/T C/P C/T +Hs +Wp STI wc50 [db] SPL(A) [db] (Left) * * τ 1 [ms] (Right) * (Left) * * * 0.35* 0.38* 0.37* φ 1 (Right) * 0.31* * * * 0.36* 0.38* 0.37* IACC * * 0.41* 0.44*

12 Fig. 9 - Temporal and spectral characteristics of PA announcements (left channel) according to dimensional alteration and various finishing materials