Acoustics for speech in classrooms and meeting rooms

Publisher’s version / Version de l'éditeur:

Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n’arrivez pas à les repérer, communiquez avec nous à PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca.

Questions? Contact the NRC Publications Archive team at

PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. If you wish to email the authors directly, please see the first page of the publication for their contact information.

https://publications-cnrc.canada.ca/fra/droits

L’accès à ce site Web et l’utilisation de son contenu sont assujettis aux conditions présentées dans le site LISEZ CES CONDITIONS ATTENTIVEMENT AVANT D’UTILISER CE SITE WEB.

WESPAC: Western Pacific Acoustics Conference [Proceedings], pp. 1-4,

2003-04-01

READ THESE TERMS AND CONDITIONS CAREFULLY BEFORE USING THIS WEBSITE.

https://nrc-publications.canada.ca/eng/copyright

NRC Publications Archive Record / Notice des Archives des publications du CNRC :

https://nrc-publications.canada.ca/eng/view/object/?id=fa3cc7ed-161c-4efa-b3ca-4959adaec3d4 https://publications-cnrc.canada.ca/fra/voir/objet/?id=fa3cc7ed-161c-4efa-b3ca-4959adaec3d4

NRC Publications Archive

Archives des publications du CNRC

This publication could be one of several versions: author’s original, accepted manuscript or the publisher’s version. / La version de cette publication peut être l’une des suivantes : la version prépublication de l’auteur, la version acceptée du manuscrit ou la version de l’éditeur.

Access and use of this website and the material on it are subject to the Terms and Conditions set forth at

Acoustics for speech in classrooms and meeting rooms

Acoustics for speech in classrooms and meeting rooms

Bradley, J.

NRCC-46120

A version of this document is published in / Une version de ce document se trouve dans :

WESPAC (Western Pacific Acoustics Conference), Melbourne, Australia,

April 7-9, 2003, pp. 1-4

Acoustics for Speech in Classrooms and Meeting Rooms

Manuscript Number: 01006G John BRADLEY

IRC, National Research Council, Montreal Rd. Ottawa, Canada K1A 0R6.

ABSTRACT

Good acoustical design should optimise room acoustics and minimise unwanted noise so that effective speech-to-noise ratios are maximised in rooms for speech. Although reported noise levels in classrooms almost always exceed ideal criteria, these results may be questioned because it is difficult to measure the speech and noise levels that actually occur during speech. Many criteria are based on studies that show a poor understanding of room acoustics and tend to prescribe more absorptive environments because they ignore the positive effects of early reflections. The more stringent requirements for various special needs groups of listeners are much less well defined. We still design rooms in terms of reverberation time that only indirectly relates to critical room acoustics details and even this we cannot do accurately. This paper reviews recent studies of these problems and outlines remaining research needs.

KEYWORDS: classroom, meeting rooms, speech in rooms, intelligibility

INTRODUCTION

The goal of good acoustical design for classrooms and meeting rooms is to make possible clear, accurate and relaxed speech communication. Listeners should be able to understand all of the words that are being spoken to them without having to strain. This is important in classrooms because the majority of teaching situations involve oral communication. In meeting rooms poor communication can lead to costly mistakes.

-10 0 10 20 0 20 40 60 80 100 S/N, dBA In te lli g ib ili ty , % 100 80 60 40 20 0 D iff ic u lty , % Difficulty Intelligibility

Fig 1. Comparison of intelligibility and difficulty ratings.

It is well known that good acoustical design requires one to maximise the signal-to-noise ratio and to provide optimum room acoustics conditions. It is common experience that speech communication is difficult in many rooms, and almost all reported noise levels in classrooms exceed recommendations, indicating that there still remain problems to be solved. The situation is complicated by the fact that speech and noise levels are

greatly influenced by room acoustics and one cannot consider the two elements in isolation.

We have quite remarkable abilities to understand speech in adverse conditions. Even when it is relatively noisy and spaces are very reverberant, we are often able, or at least we think we are able, to understand speech. In practice, we are often unaware that we are frequently guessing (sometimes correctly, sometimes not) the meaning of the speech sounds. In addition there are a number of special groups of listeners who have greater difficulty understanding speech when it is too noisy or too reverberant. These groups would include younger listeners, older listeners, second language listeners and anyone with

even a modest amount of hearing loss. This paper will review what we do know about the acoustical requirements of rooms for speech and the problems that remain to be solved.

SIGNAL-TO-NOISE RATIOS

Fig 2. Reported classroom noise levels.

20 30 40 50 60 70 80 0 50 100 150 200 250 300 350 Noise Mean 46.0 dBA Leq 47.8 dBA Speech Mean 56.5 dBA Leq 58.2 dBA F req ue nc y o f oc cu re nc e SPL, dBA

Fig 3. Distribution of speech and noise levels.

-20 -10 0 10 20 3 40 60 80 100 0 70 60 50 40 30 20 6 year olds 13 year olds Int elligibi lity, % Signal/Noise, dBA

Fig. 4. Estimated intelligibility for 6 and 13 year olds vs. S/N ratio.

The louder speech sounds are relative to interfering noises, then the greater the intelligibility of the

speech. Several studies have indicated that a signal-to-noise ratio (S/N) of ≥ +15 dBA provides

conditions in which 100% intelligibility scores are possible on a simple speech intelligibility test. There are many different types of speech intelligibility or speech recognition tests. Although the shape of the curves of speech intelligibility scores versus S/N vary among the various tests, they would all

indicate very close to 100% intelligibility at a S/N of ≥ +15 dB. However, one may still question

whether 100% on a speech intelligibility test is truly optimum, because conditions do not correspond to relaxed listening. Typically, listeners are straining to understand the test words. To solve this problem, Sato [1] has proposed using subjective ratings of the ‘Difficulty’ of listening to speech. Figure 1 compares intelligibility and difficulty ratings of the same test conditions. For S/N values greater than 0 dB, intelligibility scores gradually increase

to close to 100% intelligibility. Over the same S/N range, difficulty ratings change quite dramatically indicating that listeners noticed very large changes in conditions even though they were almost always able (with considerable strain) to understand the test words.

To maximise S/N ratios in classrooms, one should first attempt to minimise ambient noise levels. A recent review of classroom acoustics issues [2] found noise levels measured in classrooms to vary from 42 to 65 dBA and to even higher in day care facilities for pre-school children. (See Fig. 2) There is a trend for classrooms with younger children to be noisier. These results indicate noise levels greater than recommended in a recently published ANSI standard for classroom acoustics (35 dBA recommended).

It is quite difficult to get accurate measurements of speech and noise levels in classrooms that are representative of the conditions during actual speech tests. Frequently, the measured speech levels may include the influence of ambient noise and ambient noise measurements may not be representative of the conditions during speech communication. Representative speech and noise levels may be obtained from analyses of statistical distributions of speech and noise levels [3]. Figure 3 illustrates an example of such an analysis for the average of measurements at four positions for a female teacher in a high school classroom. By fitting two normal distributions to the measured data, one can explain the measured results as due to the combination of ambient noise and speech components. Using this technique indicated speech levels (@ 1m) of 65 to 70 dBA for teachers.

A number of particular groups of listeners are thought to need a S/N of greater than +15 dB to achieve 100% intelligibility. These groups would include listeners with even a modest amount of hearing impairment, but also very young and very old listeners, as well as second language listeners. Figure 4 shows an estimate of the expected intelligibility versus S/N ratio relationship for 6 year olds compared to the mean trend of measured results for 13 year-olds [2].

ROOM ACOUSTICS

Although we usually characterise room acoustics in terms of reverberation time, there is much more detail to be considered to understand how rooms influence speech perception. In a room, listeners hear the direct sound of the talker followed by many delayed reflections of the speech. Although we typically hear the combined effect of many thousands of reflections, all reflections are not equal, and they do not all affect us in the same way. Early-arriving reflections within about 50 ms after the direct sound are particularly important because our hearing system integrates them with the direct sound, making it seem louder. Thus increased early-reflection energy is expected to increase intelligibility. Later-arriving speech sounds are not integrated and cause one speech sound to blur into the next, decreasing intelligibility. Although the importance of early-arriving reflections is not widely appreciated, it is not a new concept and Joseph Henry explained the key points in the 1850s. The concept that the ratio of early-arriving to late-arriving speech sounds would relate to speech intelligibility developed from work by Thiele in the 1950s.

Recent experiments directly showed the benefit of early reflections on speech intelligibility scores [4] using an 8-channel electro-acoustic system in an anechoic room to simulate a range of conditions. Each loudspeaker reproduced one early reflection followed by a reverberant decay starting at approximately 50 ms after the direct sound. Subjects listened to speech sounds with various combinations of direct sound with early and late reflections combined with a typical ambient noise spectrum (NC40 spectrum at 47 dBA).

-8 -6 -4 -2 0 2 4 6 50 60 70 80 90 100

Direct only non-impaired Direct only impaired Direct+early non-impaired Direct+early impaired Direct+early+reverb non-impaired Direct+early+reverb impaired In tel ligi b ilit y, % Effective S/N(A), dB Fig. 5. Benefit of early reflection to

intelligibility for impaired and

non-Sound fields were first created with only a direct sound that was varied in amplitude relative to the fixed background noise. For these cases intelligibility scores increased with increasing S/N ratio. The effective speech levels were then varied by adding increasing amounts of early reflection energy to a fixed direct sound component. The results, in Figure 5, showed that adding early-reflection energy has the same effect on intelligibility scores as increasing the direct sound level. The subjects included a number of middle aged listeners with some moderate hearing impairment and the results were separated into two groups according to their hearing sensitivity. One group consisted of mostly university students and younger adults who had near-perfect hearing, while the second group was more typically middle aged with some mild to moderate hearing loss.

The separate results for the two groups of subjects in Figure 5 showed that their hearing sensitivity had a large effect on their intelligibility scores. For both groups, added early-reflection energy was equally beneficial to a similar increase in the level of the direct sound. The third sets of results in Figure 5 show that the same statistically significant pattern of results is maintained, when the simulated sound fields included reverberant decays for both groups of listeners.

The amount of early-reflection energy in real rooms was assessed in terms of the Early Reflection Benefit (ERB) which is the total energy in the first 50 ms of the impulse response relative to that in the direct sound. Calculation of the ERB values averaged over the octave bands from 1 to 4 kHz showed ERB values varied from 1 or 2 dB, at positions close to the source, to as much as 9 dB at larger distances in rooms. Thus in actual rooms the benefit from early reflections is quite substantial and at more distant positions it is often only possible to understand speech because the early-reflection energy significantly increases the effective speech levels.

CRITERIA AND DESIGN

One can make simple estimates of the maximum desirable ambient noise level from typical speech source levels, the typical attenuation across a room, the required S/N ratio, and any increased S/N for special listener groups. For example: with a speech source level of 60 dBA, a 5 dB attenuation across the room, a 15 dB S/N ratio requirement and an additional 5 dB for younger listeners would result in a

maximum acceptable noise level of 35 dBA. However, this ignores the effects of room acoustics details on speech levels. More complete derivations of criteria can be obtained from combined acoustical measures that include the effects of both room acoustics and S/N such as Useful-to-Detrimental sound ratios.

-2 -1 0 1.0 2 3 4 5 6 0.5 1.0 1.5 2.0 2.5 25 30 35 40 45 50 55 RT = 0.5 s, BG(A) = 45 dBA BG(A), d B A RT, s U50

Fig. 6. Equal U50 contours showing combinations corresponding to equal intelligibility.

100 1000 10000 20 30 40 50 60 55 dBA 60 dBA 65 dBA BG(A ), d B A Volume, m3 45 dBA 300 m3 room Voice level

Fig. 7. Maximum ambient noise levels vs. room volume and by voice level.

100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 RT , s Volume, m3 Optimum RT Knudsen & Harris

Fig. 8. Optimum reverberation time vs. room volume.

Lochner and Burger developed the concept of useful-to-detrimental sound ratios, where ‘useful’ is the combination of the direct and early-reflected sound and ‘detrimental’ is the sum of the late-arriving speech sounds plus the ambient noise. U50 is one such Useful-to-Detrimental ratio where early-reflection energy is summed over the first 50 ms. The Speech Transmission Index (STI) is a more recent measure that combines both room acoustics and S/N aspects into a single measure. Although the STI measure is quite complex and appears to be very different than the useful-to-detrimental sound ratio, the two measures are actually very closely related [5].

Fig. 6 shows calculated equal U50 contours for

positions in a 300 m3 _{room and an assumed speech}

source level of 60 dBA. A U50 of +1 dB corresponds to near perfect intelligibility. Thus points on the U50 = +1 contour represent combinations of ambient noise level and reverberation time (RT) that would all lead to the same near perfect intelligibility conditions. This contour has a maximum at approximately a 0.5 s RT and a maximum background noise level of 45 dBA. Although other points on the same contour would represent equal intelligibility conditions, the maximum point (45 dBA, RT=0.5 s) represents an optimum because the most noise can be tolerated at this point.

One can calculate similar sets of equal U50 (i.e. equal intelligibility) contours for a range of speech source levels and room volumes. From the optimum point of each contour one can derive maximum acceptable ambient noise levels and optimum reverberation times as illustrated in Fig. 7 and 8.

CONLCUSIONS

We know enough to specify reasonable estimates for

optimum conditions for speech in rooms. Obtaining acceptable ambient noise levels is usually more critical than achieving optimum reverberation times.

To be more precise about setting criteria, we need to better determine representative and safe voice levels for talkers, and the additional needs of various particular listener groups. We also need to decide whether 100 % intelligibility is adequate for relaxed listening.

Finally, we need to learn how to design rooms to maximise early reflection energy without excessive later arriving speech sounds.

References [1] Sato, Bradley and Morimoto., Canadian Acoustics. 30 (3) 50-51, (2002).

[2] Picard and Bradley, Audiology, Journal of Auditory Communication, vol. 40, no. 5, pp. 221-244, (2001). [3] Hodgson, Rempel and Kennedy, J. Acoust. Soc. Am. 105(1) 226-233 (1999).

[4] Bradley, Sato and Picard, J. Acoust. Soc. Am., 111 (5) Pt. 2, p. 2411, (2002). (accepted for pub. J. Acoust. Soc. Am.) [5] Bradley, J. Aud. Eng. Soc., Vol. 46, No. 5, pp. 396-405 (1998).