Introduction: The effect of a sound reinforcement system, in terms of speech intelligibility, has been systematically determined under realistic conditions. Different combinations of ambient and reverberant conditions representative of a classroom environment have been investigated. Materials and Methods: By comparing the measured speech transmission index metric with and without the system in the same space under different room acoustics conditions, it was possible to determine when the system was most effective. A new simple criterion, equivalent noise reduction (ENR), was introduced to determine the effectiveness of the sound reinforcement system which can be used to predict the speech transmission index based on the ambient sound pressure and reverberation time with and without amplification. Results: This criterion had a correlation, R2 > 0.97. It was found that sound reinforcement provided no benefit if the competing noise level was less than 40 dBA. However, the maximum benefit of such a system was equivalent to a 7.7 dBA noise reduction. Conclusion: Using the ENR model, it would be possible to determine the suitability of implementing sound reinforcement systems in any room, thus providing a tool to determine if natural acoustic treatment or sound field amplification would be of most benefit to the occupants of any particular room.
Keywords: Acoustics, amplification, classroom, speech
|How to cite this article:|
Dance S, Backus B, Morales L. A methodology to objectively assess the performance of sound field amplification systems demonstrated using 50 physical simulations of classroom conditions. Noise Health 2018;20:77-82
|How to cite this URL:|
Dance S, Backus B, Morales L. A methodology to objectively assess the performance of sound field amplification systems demonstrated using 50 physical simulations of classroom conditions. Noise Health [serial online] 2018 [cited 2018 Dec 15];20:77-82. Available from: http://www.noiseandhealth.org/text.asp?2018/20/94/77/232705
| Introduction|| |
The assessment of the acoustical quality of schools both objectively and subjectively has been extensively investigated.,, It has been found that speech intelligibility in rooms with poor acoustics was detrimental to learning,,, as have classrooms with high levels of background noise.,, There were particular acoustic-related learning difficulties found for bilingual children or children with additional needs in poorly performing spaces. However, sound field amplification (SFA) has been found to improve children’s behavior in cross-cultural environments.,, It has also been found that the teacher’s voice will suffer in classrooms with poor acoustics., As such, new guidance has recently been introduced to aid in the design of classrooms, particularly rooms for more than 40 students. There are now two newly available solutions: personal amplification systems which have been found to work well for school children and single speaker SFA, Phonak, which is the focus of this paper.
Previously, SFA required that the room be fitted with an array of loudspeakers and hence the system was costly. In addition, there are installation implications in that the ceiling of the classroom would need to be removed. The Phonak SFA system has the benefit over a traditional SFA system in requiring two pieces of apparatus: a microphone and a single floor standing loudspeaker array to be installed in the classroom. Hence, there are significant cost savings and greatly reduced installation implications. However, the training cost of learning to use the system would be similar.
With a significant investment in new schools and a program of refurbishment currently being undertaken in the United Kingdom under new produced guidance, it is necessary to determine how to most effectively improve speech intelligibility in classrooms, either by reducing the room reverberation through the application of acoustic treatment, the introduction of a sound reinforcement system into room, or a combination of the two approaches. This type of comparison has not been undertaken before, as the algorithms used in sound reinforcement systems are not disclosed. To this end, this paper attempts to provide a methodology to objectively quantify under what conditions such a system improves speech intelligibility considering two variables: ambient noise and reverberation both with and without the sound reinforcement system. The paper details the equipment setup, the experimental method used, and provides results in the form of a new term: effective noise reduction (ENR) for the various scenarios, 50 in total. From these results, an empirical model was proposed which could be used as a tool to help make informed decisions as to what approach to take in refurnishing classrooms or similar spaces.
| Materials and Methods|| |
The selected equipment as regards the SFA system was the Phonak Digimaster 5000 (Phonak, Warrington). It was used to demonstrate the proposed methodology and procedure to determine the effectiveness of SFA in noisy and reverberant classrooms, but could be installed in alternative spaces.
The Digimaster 5000 does automatically monitor background noise level, in, for instance, a classroom setting and adapts the gain level of the speaker’s voice dynamically. This ensures an appropriate signal-to-noise ratio for the teacher’s voice, independent of whether the children are quiet or whether there is a high noise level in the classroom. It should be mentioned that Phonak are most famous for their hearing aid designs which do contain compression algorithms. We do not know if they were implemented in the SFA system.
The reverberation chamber at London South Bank University was used for the tests to physically simulate typical classrooms, volume of 202 m3 and surface area of 213 m2, very similar to that measured by Shield et al. in 165 British secondary schools [Table 1], an average volume of 217 m3. The physically simulated classroom was then set up in a number of acoustic conditions, with a baseline background noise level of 35 dBA with an empty reverberation time (T60) 1 kHz of 5.1 s, that is, the time it takes for a sound to decay by one million times. Porous sound absorbing material (0.17 m thick hung in strips on the wall) was introduced into the space to achieve different reverberation times (T60) at 1 kHz (0.4, 0.6, 0.8, 1.0, and 1.2 s) [Figure 1]. To achieve the significant reduction in reverberation times (5.1–0.6 s), it was necessary to hang an increasing number of rolls of 170 mm deep mineral wool from the top of each wall in the reverberation chamber. To achieve T60 0.4 s, it was also necessary to position an additional roll of mineral wool on the floor of the chamber [Figure 1]. These reverberation times gave a range which was very similar to that measured by Shield et al. [Table 2]. As the reverberation chamber was designed to produce a diffuse sound field, it was assumed that the location of the absorption would not affect the results.
|Figure 1: The reverberation chamber showing the measurement setup including absorption, T60 = 0.4 s|
Click here to view
|Table 2: Summary of the acoustic condition of British Secondary School Classrooms|
Click here to view
A mouth simulator (Berhinger B205D loudspeaker) was positioned at the front of the room at a height of 1.5 m. Speech intelligibility measurements were taken using the Speech Intelligibility Index for Public Address (STIPA) systems which simulates the male voice modulations and spectral envelope. The STIPA signal was set to 65 dBLAeq,15s as measured 1 m from the loudspeaker in the free field, equivalent to a raised voice level. This signal was used to determine the speech transmission index using the STIPA method. The signal level was found to be 61.2 dBLAeq,15s in the physically simulated classroom for the most reverberant case (T60) 1 kHz of 1.2 s as measured 2.5 m from the loudspeaker. In addition, two loudspeakers (Yahama HS50M), positioned 2.5 m from the measurement microphone [Figure 2], generated competing filtered random noise with the same spectral envelope as the STIPA signal, but without modulation at five levels to give a controlled signal-to-noise ratio [Table 3].
|Figure 2: The measured frequency dependence of the reverberation times (T60) for the five different room setups (T60, 1 kHz = 0.4, 0.6, 0.8, 1.0, and 1.2)|
Click here to view
|Table 3: Measurement conditions showing the combinations of achieved reverberation times (T60 at 1 kHz) and competing noise levels used for the individual measurement trials|
Click here to view
For each T60 value, a baseline STIPA measurement was made for each competing noise level (40, 50, 55, 60, and 70 dBA) chosen [Table 3], again similar to that found by Shield et al. [Table 2]. Once the baseline was determined, the SFA boom microphone was placed on axis at a distance of 0.05 m, and the system activated and allowed to acclimatize to the noise for 20 s before the measurement was taken. On the basis of this measurement, the introduced competing noise was adjusted until the original STIPA value was again achieved (within 0.02 over an average of three 15 s measurements).
Next, the SFA system was turned off and the ambient noise level was measured. The difference between the noise levels with and without SFA that produced the same criterion STIPA was taken as a new metric, termed the equivalent noise reduction (ENR) achieved by the system.
Finally, the room impulse responses were measured under background noise level conditions and at each of the five noise levels using eSweeps in winMLS 2004 in accordance to ISO 3382-1:2009. This was used to measure and verify the physically simulated classroom’s reverberation times [Figure 3]. Because of the dynamic nature of the SFA system, it was not possible to measure impulse responses (which requires linear time invariance) when SFA system active.
|Figure 3: The position of the mouth simulator (SA), the sound field amplification system (SB), and two noise generating sound sources (NA and NB)|
Click here to view
| Results and Analysis|| |
The STIPA measurement results were plotted against competing noise levels and this data was fitted experimentally to determine the following equation:
where x represents competing noise in terms of pressure, Parms (not dB SPL). The two free variables, A and α (governing the y-axis intercept and curvature), were calculated using a Nelder–Mead search method to minimize the local sum of the squared error between the model and the data.
R2 was calculated and used to evaluate the goodness of fit between the model and measured data. Of course, Eq. (1) is only a model for the Phonak SFA system under investigation, and of course, could theoretically give an speech transmission index (STI) outside of the 0–1 range.
Activating the SFA system increased the STIPA by an amount equivalent to the increase one could achieve by reducing the competing noise [Figure 4]. The amount of this “equivalent noise reduction” or ENR was found to depend on the level of noise, the room’s reverberation time and presumably the placement of the system relative to the listener. This new simpler parameter is easier for non-acousticians to understand, as it is in the units of decibels, and it does not require a modulated signal and hence works even if the SFA system compresses the signal.
|Figure 4: The effect of adding the sound-field reinforcement system on STIPA terms of an equivalent noise reduction for five different room configurations (curves)|
Click here to view
From [Figure 4], no increase in the STI metric was observed for background noise below 40 dBA, confirming the field measurement results of Dockrell and Shield. The STI metric monotonically increased with increasing competing noise, but eventually saturated at 7.7 dBA ENR at the measurement microphone position.
The same data were replotted as the measured STI metric vs. the measured competing noise (dBA) [Figure 5]. The same two parameter model was used to fit both SFA and no-SFA datasets with excellent correlation found, all R2 values >0.97 (refer [Figure 5]A and [Figure 5]B, respectively).
|Figure 5: (A) How STIPA changed with competing noise for various room reverberation times (curves) without sound-field amplification. Data points (solid) are overlaid with a two parameter mathematical model (dashed lines). (B) The same type of plot with sound-field amplification. (C) A comparison of the model fits extracted from B|
Click here to view
| Discussion|| |
All SFA systems are Proprietary and as such, the details of how the particular system works are unknown. The methodology detailed above was designed so that a comparative performance of any SFA system could be undertaken in a room while varying reverberation and the signal to noise levels. Ideally, the proposed objective SFA system performance test should also be tested subjectively with human listeners to fully demonstrate performance, although this solution would be affected by human factors such as hearing sensitivity and auditory processing differences.
For room reverberation times ≥1.0 s, adding SFA increased STIPA values for competing noise levels above 50 dBA [[Figure 5]C]. For reverberation times ≤0.6 s, this threshold was 38 dBA and for T60 = 0.8 s, it was 44 dBA. Below these thresholds, the modeled data indicate that adding SFA would actually reduce STIPA and that this deterioration would be larger for more reverberant rooms, for example, modeled STIPA was reduced by 0.05 for T60 = 1.2 s, refer [Figure 5]C for 30 dBA background noise. So the SFA system improves intelligibility with increasing competing background noise but this improvement is mitigated by increasing reverberation. This is expected as the amplification in the SFA works well but the reverberation creates masking in the signal reducing the intelligibility.
[Table 4] shows the coefficients A and α from Eq. (1) and the correlation to the measured STIPA values based on competing background noise levels. From [Figure 6], it can be seen that both A and α coefficients were affected by activating the SFA, but only parameter A was sensitive to room reverberation time. Taken together, these parameters suggest that just by taking account of room reverberation, a simple model can capture the effects of SFA in noisy environments in terms of speech intelligibility.
|Table 4: Model parameter values and goodness of fit, R2 values, of the empirical model with SFA|
Click here to view
|Figure 6: How model parameters controlling the models curvature (parameter α) and y-intercept (parameter A) are affected by reverberation time and activating the SFA|
Click here to view
As classrooms normally have a diffuse sound field the speech intellibillity would be uniform, and as such, increasing the distance from the loudspeaker to the child would have no effect. It should be noted that Phonak recommends one unit for a small classroom and 30 students and two for a larger classroom.
| Conclusion|| |
A new methodology to evaluate the performance of SFA systems has been detailed, demonstrated using a physically simulated classroom. After a series of laboratory-based measurements in the simulated classroom, using a combination of different reverberant and competing noise levels, the ENR parameter was developed to indicate the potential benefit of SFA systems. Although the specific benefit is limited to one type of SFA, the methodology could be applied to any room-based system and any system whether linear or nonlinear based, for example, using compression algorithms.
On the basis of these measurements, an empirical model was developed to predict the expected speech intelligibility performance improvement when using a sound amplification system in noisy environments under a range of reverberant conditions. The ENR parameter provided a method by which the value (performance/cost) of adding a SFA system to a room could be compared with the more traditional room acoustic treatment solutions.
Financial support and sponsorship
Conflicts of interest
For clarity, Phonak offered no inducement to use their products. The research team had access to the Phonak system for a period of 3 months to undertake any experiment of their choosing and had the right to publish any and all of the results discovered.
| References|| |
Astolfi A, Pellerey F. Subjective and objective assessment of acoustical and overall environmental quality in secondary school classrooms. J Acoust Soc Am 2009;123:163.
Yang W, Bradley JS. Effects of room acoustic on the intelligibility of speech in classrooms for young children. J Acoust Soc Am 2009;125:922.
Astolfi A, Bottalico P, Barbato G. Subjective and objective speech intelligibility investigations in primary school classrooms. J Acoust Soc Am 2012;131:247.
Sato H, Morimoto M, Wada M. Relationship between listening difficulty rating and objective measures in reverberant and noisy sound field for young and elderly persons. J Acoust Soc Am 2012;131:4596.
Valente D, Plevinsky H, Franco J, Heinrichs-Graham E, Lewis DE. Experimental investigation of the effects of the acoustical conditions in a simulated classroom on speech recognition and learning in children. J Acoust Soc Am 2012;131:232.
Bistafa SR, Bradley JS. Reverberation time and maximum background noise level for classrooms for a comparative study of speech intelligibility metrics. J Acoust Soc Am 2000;107:861.
Houtgast T. The effect of ambient noise on speech intelligibility in classrooms. Appl Acoust 1981;14:15-25.
Shield BM, Dockrell J. The effects of environmental and classroom noise on the academic attainments of primary school children. J Acoust Soc Am 2009;123:133.
Dockrell J, Shield BM. Acoustical barriers in classrooms: The impact of noise on performance in the classroom. Br Educ Res J 2009;32:509-25.
UK Education Funding Agency. Acoustic Performance Standards for the Priority Schools Building Programme. London: Education and Skill Funding Agency; September, 2012.
Massie R, Dillon H. The impact of sound-field amplification in mainstream crosscultural classrooms: Part 1-Educational outcomes. Aust J Educ 2006;50:62-77.
Massie R, Dillon H. The impact of sound-field amplification in mainstream crosscultural classrooms: Part 2-Teacher and child opinions. Aust J Educ 2006;50:78-94.
Rosenberg G, Blake-Rahter P, Heavner J, Allen L, Redmond B, Philips J et al.
Improving classroom acoustics (ICA): A three-year FM sound field classroom amplification study. J Educ Audiol 1999;7:8-28.
Durup N, Shield B, Dance S, Sullivan R. How classroom acoustics affect the voice parameters of teachers. J Build Acoust 2015;22:225-42.
Lyberg-Åhlander V, Rydell R, Löfqvist A. Speaker’s comfort in teaching environments: Voice problems in Swedish teaching staff. J Voice 2011;25:430-40.
Pelegrín-García J, Brunskog B, Rasmussen B. Speaker-oriented classroom acoustics design guidelines in the context of current regulations in European countries. Acta Acust United Acust 2014;100:1073-89.
Rawool V, editor. Auditory Processing Deficits: Assessment and Intervention. NY, USA: Thieme; 2015. ISBN 9781604068375.
Phonak Dynamic Sound Field User Manual. Switzerland: Phonak, Stafa; 2015.
ANSI Standard ANSI/ASA S12.60-2010/Part 1 American National Standard Acoustical Performance Criteria, Design Requirements, and Guidelines for Schools, Part 1: Permanent Schools. USA: Acoustical Society of America; 2010.
Shield BM, Conetta R, Dockrell J, Connolly D, Cox T, Mydlarz C. A survey of acoustic conditions and noise levels in secondary school classrooms in England. J Acoust Soc Am 2015;137:177-88.
IEC. Sound System Equipment. Part 16: Objective Rating of Speech Intelligibility by Speech Transmission Index. 4th
ed. IEC; 2011. International Standard No. 60268-16. Geneva, Switzerland; 2004.
International Standards Organisation 3382-1:2009. Acoustics − Measurement of Room Acoustics Parameters. Part 1: Performance Spaces. Geneva, Switzerland: ASHA.org.
Dockrell J, Shield BM. The impact of sound field systems on learning and attention in elementary school classrooms. J Speech Lang Hear Res 2012;55:1163-76.
School of the Built Environment, Borough Road, London South Bank University, London SE1 0AA, London
Source of Support: None, Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2], [Table 3], [Table 4]