Predicting classroom acoustical parameters for occupied conditions from unoccupied data

Predicting classroom acoustical parameters for occupied conditions from unoccupied data

Applied Acoustics 127 (2017) 89–94 Contents lists available at ScienceDirect Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust P...

699KB Sizes 0 Downloads 61 Views

Applied Acoustics 127 (2017) 89–94

Contents lists available at ScienceDirect

Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust

Predicting classroom acoustical parameters for occupied conditions from unoccupied data Young-Ji Choi Department of Architectural Engineering, Kangwon National University, 1 Kangwondaehak-gil, Choncheon-si, Kangwon-do 200-701, South Korea

a r t i c l e

i n f o

Article history: Received 24 October 2016 Received in revised form 22 April 2017 Accepted 31 May 2017

Keywords: Total room sound absorption Occupants Acoustical parameters Classrooms Speech intelligibility

a b s t r a c t This paper introduces a simple procedure for estimating the effects of occupants on the acoustical characteristics of university classrooms. Values of room acoustics parameters such as EDT, C50, G, Glate, STI and U50 are useful indicators of the quality of conditions for speech in rooms intended for speech communication. It is more difficult to measure these parameters in occupied spaces and hence being able to predict occupied values of these acoustical parameters from unoccupied measurements is a great asset to achieving acoustically successful classrooms. In this work it is shown that the changes at 1000 Hz of EDT, C50, Glate, and U50 values, due to adding occupants, can be linearly related to the corresponding changes in total room sound absorption at 1000 Hz. It has also been shown that changes in total room absorption due to adding occupants can be simply related to the total unoccupied room absorption. Combining these two relationships provides a simple procedure for predicting the values at 1000 Hz of EDT, C50, Glate, and U50 values in occupied classrooms. These new procedures make it possible to design better classroom acoustics and to better understand acoustical problems when they occur. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction A number of studies [1–3] have reported the mean sound absorption per occupant determined from the measured reverberation times in classrooms. Table 1 compares the mean sound absorption per occupant in classrooms from the results in several previous studies [1–3]. As shown in Table 1, the mean sound absorption per occupant varies with frequency, chair type, classroom design, and the ages of the occupants [1–4]. That is, the changes to the values of room acoustics parameters due to added occupants in classrooms are influenced by the percentage change of the total sound absorption [3] and also by the perimeter-toarea ratio of each block of occupied chairs [3,5]. Two university classroom acoustical studies [1,3] have reported that the effect of the added absorption of occupants is dependent on the acoustical conditions of the classroom. That is, the addition of occupants to the more reflective classrooms led to larger fractional increases in the total sound absorption of the classrooms and to larger incremental changes to the values of the related room acoustics parameters [1,3]. The changes to the values of room acoustics parameters were: to the perceived reverberance as measured by EDT values, to the clarity of speech sounds indicated by C50 values, to the overall levels and later-arriving reflection levels E-mail address: [email protected] http://dx.doi.org/10.1016/j.apacoust.2017.05.036 0003-682X/Ó 2017 Elsevier Ltd. All rights reserved.

resulting from sound strength indicated by G values and Glate values, and to the speech intelligibility as measured by STI values. The results of the two university classroom acoustics studies [1,3] also showed a significant effect of the presence of occupants on the acoustical conditions in classrooms, emphasizing the need for design criteria for occupied classrooms. The results of the speech intelligibility predictions for 279 university classrooms [6] showed that the acoustical quality determined by the room-average speech intelligibility of classrooms decreases with decreasing occupancy. A more recent study in 12 university classrooms [3] concluded that added occupants may contribute to achieving more ideal reverberation times for speech (typically 0.4–0.7 s7) in the more reflective classrooms, but not in the more absorptive classrooms. This paper is a follow-up to previous work [3] that experimentally investigated the effect of occupancy on acoustical conditions in university classrooms. In the previous work [3], a simple process for estimating the added sound absorption per occupant to classrooms that are similar to the 12 university classrooms was proposed and described by Eq. (1) (adopted from Eq. (3) in Ref. [3]). Linear regression lines were fitted to a plot of the total measured sound absorption of the occupied classrooms versus that of the unoccupied classrooms at mid-frequencies (500 Hz and 1 kHz octave bands) and the regression coefficients for the plotted regression lines are included in Eq. (1). One could estimate average

90

Y.-J. Choi / Applied Acoustics 127 (2017) 89–94

Table 1 Mean sound absorption per occupant (m2/person) determined from the measured reverberation times in the classrooms reported in previous studies [1–3]. Classrooms (researcher, year)

Frequency, Hz

10 university classrooms [1] 30 elementary classrooms [2] 6 reflective university classrooms [3] 6 absorptive university classrooms [3]

125

250

500

1000

2000

4000

– 0.08 0.18 0.31

0.45 0.10 0.33 0.34

0.67 0.23 0.60 0.63

0.81 0.46 0.79 0.45

0.83 0.41 0.80 0.33

0.84 0.31 0.84 0.26

octave band values (500 Hz and 1 kHz octave bands) of the sound absorption added per occupant in classrooms and apply these values to other classrooms. One might expect classrooms close to 54% occupancy to be best predicted by Eq. (1).

Aocc ¼ 1:05  Aunocc þ 18:5; m2 ðR2 ¼ 0:987;

r ¼ 0:038 m2 Þ

Table 3 Mean 1000 Hz total sound absorption for occupied and unoccupied classrooms and the ratios of occupied-to-unoccupied total room sound absorption. Classrooms

ð1Þ

2

where Aocc = occupied absorption in m , and Aunocc = unoccupied absorption in m2. Although this simple equation would give reasonable estimates of the occupied sound absorption values in classrooms, it would be more helpful to develop procedures for predicting the expected changes to the values of room acoustics parameters. This process is based on the effects of the values of added absorption per person on the unoccupied total absorption values of the classrooms and on other room acoustics parameters. The goal of the present work is to propose a simple method to use the added absorption per person values to predict the expected values of the acoustical parameters in occupied classrooms.

#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12

Mean 1000 Hz total sound absorption, m2 Occupied

Unoccupied

35.5 42.1 59.9 48.4 72.0 53.0 381.8 313.6 415.0 184.5 454.7 258.3

23.1 35.5 40.5 33.8 31.2 22.9 377.0 245.1 388.7 153.2 428.8 242.9

The ratio of occupied-to-unoccupied total room sound absorption

1.53 1.19 1.48 1.43 2.30 2.31 1.01 1.28 1.07 1.20 1.06 1.06

2.2. Calculation of incremental changes to the values of room acoustics parameters

2. Measurement data 2.1. Calculation of ratios of occupied-to-unoccupied total room sound absorption In this section, the measured data for 12 classrooms in previous paper Ref. [3] were used for calculating the total sound absorption of occupied and unoccupied classrooms. Table 2 presents the mean dimensions of the 12 university classrooms and the mean percentage of seats occupied during the measurements. More details of the classroom measurements are included in Ref. [3]. Table 3 presents the mean 1000 Hz total sound absorption values for occupied and unoccupied classrooms determined from the measured occupied and unoccupied reverberation times using Sabine Eq. (2).

A ¼ 0:16 V=T 30

ð2Þ 2

where, A is total sound absorption in m , V is the room volume in m3, and T30 is the reverberation time in s. These values are used to calculate the ratios of occupied-tounoccupied total room sound absorption at 1000 Hz for each classroom and are also included in Table 3. Here results were used the mean measured 1000 Hz values because the optimum reverberation times for achieving high intelligibility of speech is given values at 1000 Hz in occupied classrooms [7]. The ratios of occupied-tounoccupied total room sound absorption varied from about 1.0– 2.3 for the 12 classrooms.

The measured data for 12 classrooms in previous paper Ref. [3] were used to calculate the incremental changes to the values of the acoustical parameters, EDT, G, Glate, and C50 values, as well as in STI and U50 values, due to adding occupants to the classrooms. These data are the differences between the classrooms with and without added occupants. Useful-to-detrimental sound ratios were calculated from the measured C50 values and speech and noise levels [8]. The octave band energy ratios were weighted with the same frequency weightings as used in the STI measure [9] before summing to give the overall U50 values. For all rooms an ideal talker was assumed, to be located at the position of the sound source and speaking with a ‘Raised voice level’ according to that specified in ANSI S3.5 [10]. The expected speech levels at each receiver position were calculated assuming the source level at 1 m from the source was the ANSI ‘Raised voice level’ [10] and corresponding source spectrum. The expected attenuation to each receiver position was calculated from the measured G values. The correct expected speech levels at each receiver position were calculated using the following Eq. (3).

Lrs ¼ Lss  Attenuation ¼ Lss þ G  20; dB

ð3Þ

where, Attenuation is a positive value representing the reduction in level from a distance of 1 m to a distance of r m from the speech source. Lss is the direct speech sound level 1 m from the source, and Lrs is the speech sound level at the receiver position.

Table 2 Mean dimensions of the 12 classrooms and the mean percentage of seats occupied during the measurements.

Mean s.d. Max Min

Width, m

Depth, m

Height, m

Volume, m3

Percentage of seats occupied, %

Number of occupants

11.7 4.4 17.5 6.4

13.6 4.3 21.1 7.1

3.6 1.5 7.3 2.5

699.3 706.8 2535 193

54 30 100 20

47 27 84 11

Y.-J. Choi / Applied Acoustics 127 (2017) 89–94

91

3. Prediction of acoustical parameter increments due to adding occupants The addition of occupants is expected to absorb more reflected sound and possibly to scatter more sound in the classrooms. The sound absorption contributed by added occupants led mostly to decreased reverberation times and later-arriving reflected sounds and these effects were largest in the more reflective classrooms. That is, added occupants mostly reduce late-arriving speech levels more than early-arriving speech levels. The results from the previous study [3] clearly showed that the changes in the late arriving sound levels, i.e. Glate values, due to added occupants are large enough to be perceived as decreased reverberance. However, the smaller changes in early-arriving sound levels, i.e. G50 values, may not be perceived. Therefore, the predictions of incremental changes to G50 values due to adding occupants were not considered in the present study. Added occupants also have the effect of increasing the ratio of early-to-late energy ratios (C50), which are perceived as higher speech clarity. The C50 values have been used as a measure of the effects of room acoustics on the clarity of speech sounds in classrooms. Of course, this measure cannot estimate the combined effects of room acoustics and background noise. Previous studies [7,8] have shown that room acoustics and signal-to-noise-ratios (SNR) both influence speech intelligibility in classrooms and hence the combined effects of room acoustics and SNR on speech intelligibility should be measured. Useful-to-detrimental sound ratio values (U50) can be determined from both signal-to-noise ratios and C50 values. The U50 measure can explain the combined effects of room acoustics and SNR values on the resulting speech intelligibility. It is intended to correctly include the balance of the importance of the SNR and the room acoustic clarity. Among three types of combined measures (U50, ALcons, STI) for speech intelligibility, U50 was found to be the most accurate predictor and explained 97% of the variance in speech intelligibility scores [8]. This is the main reason for further exploring the merits of using U50 in this study. However, for U50 there is no standard procedure for combining information at different frequencies or for determining the relative importance of signal-to-noise and room acoustics components. For the purpose of demonstrating their usefulness, U50 values were calculated by averaging octave band values from 125 to 4000 Hz and using the frequency weightings from the STI measure [9]. The effects of added occupants on speech intelligibility in classrooms were compared using the two types of combined measures, STI and U50 values. This section compares incremental changes to the values of the acoustical parameters, EDT, G, Glate, and C50 values, as well as in STI and U50 values due to the added absorption of occupants. Linear regression analyses were performed to each plot of incremental changes on acoustical parameter values versus fractional increase in the total sound absorption of rooms with added occupants. The two classrooms #5 and #6 in Table 3 with the highest ratios of occupied-to-unoccupied total sound absorption at 1000 Hz (2.30 and 2.31) deviate more significantly from the trends of the other classrooms with smaller ratios of occupied-to-unoccupied room absorption values. These two classrooms were not included in the regression analyses illustrated in Figs. 1–4. The regression results calculated from the measurements of 10 classrooms are included in Table 4. Significant regression results (p < 0.05) are found for the incremental changes to EDT, C50, Glate and U50 values, but not for the changes to G, and STI values (see Table 4). Figs. 1–4 plot the measured changes at 1000 Hz of EDT, C50, Glate and U50 values due to adding occupants to 10 classrooms versus the ratios of occupied-to-unoccupied total sound absorption at 1000 Hz for 10 classrooms. Because the regression results in Table 4

Fig. 1. Measured incremental changes to EDT values at 1000 Hz due to adding occupants to each classroom versus the ratio of occupied-to-unoccupied total room sound absorption at 1000 Hz.

Fig. 2. Measured incremental changes to C50 values at 1000 Hz due to adding occupants to each classroom versus the ratio of occupied-to-unoccupied total room sound absorption at 1000 Hz.

Fig. 3. Measured incremental changes to Glate values at 1000 Hz due to adding occupants to each classroom versus the ratio of occupied-to-unoccupied total room sound absorption at 1000 Hz.

92

Y.-J. Choi / Applied Acoustics 127 (2017) 89–94

Fig. 4. Measured incremental changes to U50 values averaged of the octave bands 125–4000 Hz due to adding occupants to each classroom versus the ratio of occupied-to-unoccupied total room sound absorption at 1000 Hz.

show that these relationships are likely to be linear, the incremental changes in other classrooms due to adding occupants can be estimated from the resulting regression equations. The predicted changes at 1000 Hz of EDT, C50, Glate and U50 values for classroom #6 were estimated from the resulting regression equations and plotted in Figs. 1–4. The measured values for classroom #6 were also included in Figs. 1–4. By predicting these values for classroom #6, it will be demonstrating the usefulness of this simple method that uses the added absorption per person values to predict the expected values of the acoustical parameters in classrooms. These plots could be used to predict expected changes to acoustical parameter values in other classrooms based on the total sound absorption of unoccupied rooms and estimates of the added absorption due to the occupants. The results in Fig. 1 for incremental EDT values at 1000 Hz show very different effects of adding occupants to the 10 classrooms. The measured and predicted incremental changes in EDT values at 1000 Hz for classroom #6 were also included in Fig. 1. The measured and predicted increment in EDT values for the reflective classroom #6 were 1.02 s and 1.01 s, respectively. The difference between measured and predicted values was less than 1 just noticeable differences (JND), i.e. 0.05 s [11]. In spite of not including two classrooms having the highest ratios of occupied-tounoccupied total sound absorption at 1000 Hz (2.30 and 2.31) in the regression analyses in Table 4, the resulting regression line in Fig. 1 does appear to give reasonable estimates of the incremental changes in EDT values for classroom #6 due to adding occupants. The incremental changes in EDT values at 1000 Hz for the 6 more absorbing classrooms, with and without occupants, were less than 2 JND. For the 4 reflective classrooms, the magnitudes of the changes in EDT values at 1000 Hz rapidly increase with increasing ratios of occupied-to-unoccupied total room absorptions. These

effects may be larger for increased occupancy rates in the reflective classrooms. Two classrooms #6 and #10 in Table 3 having similar volumes (238 m3 and 226 m3) and seating capacities (46 and 48 seats) are identified for comparisons in Figs. 1–4. Classroom #6 is a typical lecture room with surfaces made of highly reflective materials. Classroom #10 is a teleconference room treated with porous type absorbing materials. Table 5 compares the mean 1000 Hz: EDT, C50, Glate values, as well as U50 values, averaged over all receiver positions, for the reflective classroom #6 and the absorptive classroom #10 without occupants. Comparing the results for the two classrooms with added occupants on the same type of metal chairs helps to provide insight into the effects of occupants in the different classrooms. For the classrooms #6 and #10 with 100% occupancy, the ratio of occupied-to-unoccupied total room sound absorption at 1000 Hz is 2.31 and 1.20, respectively. That is, the fractional increases in the total absorption due to adding occupants are greatest (131%) for classroom #6. For absorptive classroom 10, adding occupants caused only a 20% increase in the total sound absorption of room. The incremental effects of adding occupants on the same type of metal chairs are quite different if the classrooms have different reverberation times and hence different absorption characteristics. It is the fractional increase in the total sound absorption in the room due to the added occupants, which determines the magnitude of the changes in acoustical characteristics. The results in Fig. 2 for incremental C50 values at 1000 Hz clearly show that added occupants in the reflective classrooms lead to higher clarity values. The measured and predicted increment in C50 values for the reflective classroom #6 were 6.5 dB and 5.6 dB, respectively. The difference between measured and predicted values was less than 1 JND, i.e. 1.1 dB [8]. The resulting incremental changes in C50 values are quite different for classrooms #6 and #10 as found in EDT values in Fig. 1. For the absorptive classroom #10, the increment in C50 values at 1000 Hz due to added occupants averaged about 1 dB, which is less than 1 JND [8]. However, the measured increment in C50 values for the reflective classroom #6 was 6.5 dB, which is about 5 JND. Fig. 3 plots the measured incremental changes to Glate values at 1000 Hz due to adding occupants to each classroom versus the ratio of total sound absorption of occupied-to-unoccupied rooms at 1000 Hz. The measured and predicted increment in Glate values for the reflective classroom #6 were 7.6 dB and 6.9 dB, respectively. The difference between measured and predicted values was 0.7 dB. Unfortunately, the JND value for Glate values is not known. In Fig. 3, adding occupants to the reflective classrooms largely decreases Glate values and the magnitudes of the changes increase with increasing ratio of occupied-to-unoccupied total room sound absorptions. This is mainly due to the reduction in room reverberation times as shown in Fig. 1 and the corresponding reductions of late-arriving sound levels, Glate, with the added sound absorption of the occupants. For the absorptive classroom #10, the increments in Glate values at 1000 Hz due to added occupants averaged about 3.1 dB. However, the changes to Glate values for the reflective classroom #6 were 7.6 dB.

Table 4 Summary of: the correlation coefficients (R2), the statistical significance (p-value) of the correlation coefficients (* p < 0.2, ** p < 0.05, *** p < 0.01), the values of slopes (b), and intercepts (a) for incremental changes in rooms. N

10

Incremental changes to acoustical parameter values

2

R p b

a

DEDT

DC50

DG

DGlate

DSTI

DU50

0.879 *** 0.811 0.859

0.772 *** 4.380 4.511

0.326 ns 2.171 1.101

0.610 ** 4.557 3.554

0.161 ns 0.068 0.091

0.650 *** 0.742 0.819

93

Y.-J. Choi / Applied Acoustics 127 (2017) 89–94 Table 5 Mean 1000 Hz EDT, C50, Glate values, as well as U50 values averaged over all receiver positions for both the reflective classroom #6 and the absorptive classroom #10 without occupants. Classrooms

1000 Hz EDT, s

1000 Hz C50, dB

1000 Hz Glate, dB

125–4 kHz U50, dB

A B

1.71 0.21

2.8 14.5

22.9 0.7

0.4 2.0

Fig. 4 plots the incremental changes to the values of the combined measures, U50 due to adding occupants to each classroom versus the ratio of occupied-to-unoccupied total room absorptions at 1000 Hz. The measured and predicted increment in U50 values for the reflective classroom #6 were 1.0 dB and 0.9 dB, respectively. The difference between measured and predicted values was 0.1 dB. The regression results in Table 4 clearly shows a significant correlation between the increment in U50 values and the ratios of the occupied-to-unoccupied total sound absorptions of the rooms, than that for STI values. The resulting incremental changes in U50 values due to added occupants in the two classrooms #6 and #10 were 1 dB (increased from 0.4 to 0.6 dB) and 0.1 dB (increased from 2.0 to 2.1 dB), respectively. Unfortunately, there is no standard procedure for quality classification applied to U50 values. Nijs and Rychtáriková [12] proposed a conversion from measured STI values to quality classifications applied to U50 values. That is, classifications of STI in 0.15 steps from ‘bad’ to ‘excellent’ were used for U50 in 5 dB steps. But they did not use the frequency weightings from the STI measure [9] for calculating U50 values. When the mean STI values are plotted versus the mean frequency-weighted U50 values for the 10 occupied classrooms in Fig. 5, they show a very good fit to the linear regression line with a very small amount of scatter (R2 = 0.971 and the standard deviation about the regression line of, r = 0.007). Fig. 5 shows a 0.114 increase of STI values per 1 dB increase of U50 values. That is, both measures provide approximately the same information and one can use the linear regression in Fig. 5 to convert values of one measure to values of the other measure. The results in Fig. 4 show that, when the classrooms have higher occupied-to-unoccupied room absorption ratios, they have larger increases in U50 values due to added occupants. That is, the effects of room acoustics such as reverberance and clarity are more predominant for achieving good acoustics for speech in these

Table 6 Predicted mean 1000 Hz EDT, C50, Glate values, as well as U50 values with their standard deviations in the two classrooms #6 and #10 for 50% occupancy. Classrooms

1000 Hz EDT, s

1000 Hz C50, dB

1000 Hz Glate, dB

125–4 kHz U50, dB

#6 #10 s.d.

1.28 0.18 0.106

0.1 14.8 0.840

18.9 0.8 1.300

0.0 2.0 0.193

reflective classrooms. However, if the classrooms have more ideal reverberation times for speech, the SNR component is more critical for obtaining close-to-optimum conditions. The increased noise for these classrooms leads to lower U50 (125–4000) values and they deviate the most below the regression line indicating lower SNR values for the main trend. 4. Predicting acoustical conditions in occupied classrooms The new results in this paper provide a simple procedure for estimating the acoustical conditions in an occupied classroom from measurements in the unoccupied classroom. The total sound absorption in an unoccupied classroom can be determined from measured reverberation times in the room and using the Sabine equation to calculate the total sound absorption in the room. Other acoustical parameters should also be measured in the unoccupied classroom. Occupied values of acoustical parameters can be calculated from unoccupied values and the change in total room absorption due to adding occupants. The expected occupied absorption can be estimated using Eq. (1) which was obtained in a previous study [3] by fitting a linear regression line to a plot of total occupied absorption versus the corresponding unoccupied total absorption values measured in 12 different university classrooms. This relationship was quite accurate (R2 = 0.987, r = 0.038 m2) suggesting that for similar rooms with similar acoustical characteristics, one should be able to use it to predict expected occupied total sound absorption values from measured unoccupied absorption values. From the measured unoccupied total sound absorption values and the estimated occupied total sound absorption values one can calculate the ratios of occupied-to-unoccupied sound absorption. These ratios are used to predict increments in the values of acoustical parameters that result when occupants are added to the rooms. Since this procedure is based on the average trends of results from 12 different classrooms, it is likely to be reasonably accurate for generally similar rooms to the 12 classrooms in the original study [3]. Similar means that they should have room volumes close to the range of room volumes in the previous study, they should be similarly furnished, and the occupancy of the seats should be somewhat similar to the 54% average value found for the 12 classrooms. To illustrate this procedure Table 6 shows the predicted mean 1000 Hz values of room acoustics parameters with their standard deviations for the two example classrooms #6 (more reflective) and #10 (more absorptive) both with 50% occupancy. 5. Conclusions

Fig. 5. Calculated STI values versus mean frequency-weighted U50 values from 125 Hz to 4000 Hz for the 10 occupied classrooms.

This work has developed a simple procedure for estimating the acoustical characteristics of occupied classrooms from measurements of the unoccupied classrooms. It has shown that the incremental changes to acoustical parameters, EDT, C50, Glate, and U50, due to adding occupants to the classrooms can be predicted based on the ratio of occupied to unoccupied total sound absorption values for the rooms using the regression equations in Table 4. Table 6

94

Y.-J. Choi / Applied Acoustics 127 (2017) 89–94

showed example predictions of mean 1000 Hz values of room acoustics parameters with their standard deviations for the two example classrooms #6 and #10 with 50% occupancy. If one knows the total absorption of an unoccupied room, the total absorption when occupied can be predicted from Eq. (1). Then, the added total sound absorption of occupants can be obtained from the differences between the total absorptions with and without occupants. The expected changes in acoustical parameters due to adding occupants can be predicted using the regression coefficients given in Table 4. Because the resulting regression equations are based on the ratios of occupied-tounoccupied room absorption values, one can also apply them to predict the acoustical conditions of classrooms with varied occupancy. For example, at 1000 Hz, the measured added total sound absorption of occupants to the two classrooms with 100% occupancy was 30.1 m2 for classroom #6 and 31. 3 m2 for classroom B. Because the seating capacities for the two classrooms #6 and #10 are 46 and 48 seats, the added sound absorption per occupant at 1000 Hz is about 0.65 m2. For the two classrooms A and B with 50% occupancy, the ratio of total sound absorption of occupied-tounoccupied rooms would be 1.66 and 1.10, respectively. If the values of room acoustics parameters are known, one can predict the incremental changes to each parameter and give their room acoustical parameter values as shown in Table 6. While this paper has proposed a simple method that uses the added absorption per person values to predict the expected values of the acoustical parameters in classrooms, different arrangements and distributions of occupants in seats and different percentage occupancies may change the absorptive effects of occupants in the classrooms. The effects of these variables are not fully understood and need further investigation. Predicting the effects of changed occupancy can also be influenced by the perimeter-toarea ratio of groups of chairs. Considering these effects on the prediction of occupied acoustical conditions should also be included in future research studies. The results illustrate that useful-to-detrimental sound ratios (U50) can be used to measure the combined effects of room acoustics (C50) and SNR values on the intelligibility of speech in classrooms essentially as accurately as using STI values. However, the details of a standardized procedure for including how to combine information at different frequencies to obtain U50 values should be developed and evaluated from a wide range of conditions in real

classrooms. Because the U50 measure is based on the same basic concepts and can be calculated from commonly measured parameters (e.g. C50 and SNR values), it can be a more practically useful means of assessing and understanding room acoustics conditions for speech. Acknowledgements I am very grateful to undergraduate students in the architectural engineering course at Kangwon National University who volunteered to be occupants for the classroom measurements. I wish to acknowledge Dr. John S. Bradley at National Research Council in Canada for his advice and invaluable discussions with the preparation of the manuscript. I also wish to acknowledge the reviewer for his helpful comments with the regression analyses in Section 3. This work was supported by National Research Foundation of Korea Grant funded by the Korean Government (2015R1D1A1A01056575), and 2016 Research Grant from Kangwon National University (No. 520160492). References [1] Hodgson M. Experimental investigation of the acoustical characteristics of university classrooms. J Acoust Soc Am 1999;106:1810–9. [2] Sato H, Bradley JS. Evaluation of acoustical conditions for speech communication in working elementary school classrooms. J Acoust Soc Am 2008;123:2064–77. [3] Choi YJ. Effect of occupancy on acoustical conditions in university classrooms. Appl Acoust 2016;114:36–43. [4] Kousaie K, Hodgson M. Measurement of sound absorption of unoccupied and occupied chairs in classrooms. Can Undergrad Phys J 2002;1:7–10. [5] Bradley JS, Choi YJ, Jeong DU. Understanding chair absorption characteristics using the perimeter-to-area method. Appl Acoust 2013;74:1060–8. [6] Hodgson M. Rating, ranking, understanding acoustical quality in university classrooms. J Acoust Soc Am 2002;112:568–75. [7] Yang W, Bradley JS. Effects of room acoustics on the intelligibility of speech in classrooms for young children. J Acoust Soc Am 2009;125:922–33. [8] Bradley JS, Reich R, Norcross SG. On the combined effects of signal-to-noise ratio and room acoustics on speech intelligibility. J Acoust Soc Am 1999;106:1820–8. [9] IEC 60268–16 Edition 4.0 Sound system equipment. Part 16: objective rating of speech intelligibility by speech transmission index; 2011. [10] American National Standards Institute. American National Standard Methods for calculation of the speech intelligibility index. ANSI S3.5. New York; 1997. [11] ISO 3382-Acoustics. Measurement of the reverberation time of rooms with reference to other acoustical parameters; 2003. [12] Nijs L, Rychtáriková M. Calculating the optimum reverberation time and absorption coefficient for good speech intelligibility in classroom design using U50. Acta Acust United Acust 2011;97:93–102.