Proceedings of the Institute of Acoustics

 

 

A basic protocol to characterize classroom acoustics of primary schools

 

Arianna Astolfi¹, Politecnico di Torino, Torino, Italy
Greta Minelli², Politecnico di Torino, Torino, Italy
Giuseppina Emma Puglisi³, Politecnico di Torino, Torino, Italy

 

ABSTRACT

 

With the aim to promote a fast and effective characterization of the sound environment in educational facilities and an adequate classroom acoustics design, this work provides a basic measurement protocol which consists of a minimum number of parameters and positions to be considered. The present study involved 29 primary-school classrooms where background noise level during silent and group activities, reverberation time, speech clarity, useful-to-detrimental ratio and speech levels have been acquired in occupied condition along the main axis and in one or two offset positions. Two cut-off values of maximum reverberation time to ensure optimal acoustic conditions in the case of moderate and severe requirements, respectively, were assumed equal to 0.8 s and 0.6 s, according to literature and subjective data. For each cut-off value, classrooms were divided in two consistent groups either if they were compliant or non-compliant with such requirements, respectively. Given the strong correlation among the measured quantities, cut-off values were also identified for the other acoustical parameters. The main result of the work suggests that more convenient parameters, such as clarity in the central location of the classroom, can be used beyond reverberation time, which implies a more laborious measurement procedure.

 

1. INTRODUCTION

 

The scope of the present work has been to identify the basic parameters and their optimal values necessary to fast and effectively characterize the acoustic quality of parallelepipedal shaped primary school classrooms with small and medium volumes, i.e., between 100 and 290 m³. The study arises from the awareness that many of the available indexes and parameters used for classroom acoustic characterization are closely correlated. On the basis of this knowledge, which is well-supported by literature, there is a need to advance such characterization practices by simplifying the measurement protocols in order to save time and to allow wide experimental campaigns to be made to certify a higher number of classrooms. To this aim, 29 primary classrooms in Turin, Italy, which are heterogeneous in terms of building typologies and acoustic conditions, have been considered in this study. The database we used in Astolfi et al. (2019) [1], has been enlarged in this work, although we have used the same protocol.
 

2. MATERIALS AND METHODS

 

Measurement campaign were carried out in 29 first-grade classrooms belonging to 13 primary schools in Turin, where about 550 pupils aged from 6 to 7 years participated in this study over three scholastic years, that is, from 2016 to 2019.

 

2.1. Schools and Classrooms

 

The location, periods of construction and architectural features of the 29 schools involved in the present study differ. Moreover, they were built between 1846 and 1975, and are scattered over the city of Turin, in neighborhoods characterized by low or medium volumes of traffic [2]. The classroom volumes vary between 105 m³and 290 m³, and all the classrooms, except one (G1), have a rectangular shape. The average height of the classrooms ranges from 3.0 m to 5.3 m, and the ceilings are flat or vaulted. The floor finishes are venetian tiles, except for in one case (H1), which is instead in linoleum. The furniture consists of desks and chairs, bookshelves and blackboards. Some classrooms have undergone acoustic correction interventions.

 

2.2. Characterization of the Classroom Acoustics

 

We performed the acoustic measurements in one-day in each school in the last two months of the school year, over three scholastic years, that is, from 2017 to 2019. We carried out the measurements under occupied conditions, with an average number of 19 children per class. We also measured the reverberation time under empty conditions, at the end of each session.

 

The pupils were seated in the traditional teaching layout (in rows facing the teacher’s desk) in all the classrooms, with three or four rows of desks in each classroom, which were sometimes joined. The teacher’s desk was parallel to one of the shorter walls in each room, except for in E2, where the teacher’s desk and the blackboard were parallel to the longer side of the room, for teaching purposes. We identified the main axis in each classroom,starting from the center of the wall behind the teacher’s area.

 

Details of the measurements, set-up and equipment can be found in Astolfi et al. (2019) [1]. The parameters that were measured are: T20, which is the reverberation time when children are in a class room; T20_{_e}, which refers to the empty conditions; L_{N_sil}, which is the noise acquired when the stu dents are silent and L_{N_gr}, which is the noise acquired while the students are performing group activities. Moreover, the speech level was recorded at 1 m from the source, L_{S_ref}, and then in the other positions, and its slope per double distance along the main axis, mL_S, was evaluated; speech clarity and the useful-to-detrimental ratio referred to the mean distributions in the classroom and were la beled with the subscript “M”, e.g., C50_{_M} and U50_{_M}, or referred to single values in the center of the classroom, which were labeled with the subscript “ctr” , e.g., C50_{_ctr} and U50_{_ctr}.

 

2.3. Statistical Analysis

 

The statistical analysis was carried out with SPSS (IBM Statistics 20, IBM, Armonk, NY, the USA). Once we had checked that the distribution of the database was non-normal, non-parametric methods were used to analyze the data. We investigated any correlations between the acoustic parameters by means of Spearman’s rho [3], and further analyzed those with a p-value < 0.01 through linear regression analysis.

 

We then divided the classrooms into two groups, starting from a reference T20 value chosen arbitrarily from among the ones identified as thresholds for evaluating the compliance or non-compliance of the classroom acoustic conditions with the literature or with the considered standards. In this way, the classroom acoustics were classified as compliant (C) or non-compliant (NC) with the requirement. The Mann–Whitney U Test (MWU) was used to assess the significance of the differences between the values of the parameters in the two groups. Only the parameters with a p-value below 0.05 were considered significantly different between the two groups.

 

Once subdivided the classrooms between the two groups on the basis of the reverberation time thresholds, we resort to the receiver operating characteristic (ROC) approach three times. Firstly, the area under the curve (AUC) of each of the other acoustical parameter was inspected to assess its overall summary accuracy to classify cases between the groups of classrooms compliant (C) or non-compliant (NC) with the threshold. Secondly, the analysis of the ROC curves allowed for the identification of the most suitable threshold value for the other investigated acoustical parameters, based on the Minimum Squared Distance (MSqD) [4]. Thirdly, the accuracy and precision of the thresholds identified through the MSqD from the ROC curve were then tested to assess their ability to classify the acoustic quality of the classrooms as either C or NC [5]. In order to perform these analyses, the values of the parameters were indicated as follows:

• True Positive (TP), when the classification was C and the value of the specific parameter was correctly identified in the C group;
• True Negative (TN), when the classification was NC and the value of the specific parameter was correctly identified in the NC group;
• False Positive (FP), when the classification was C and the value of the specific parameter was identified in the NC group;
• False Negative (FN), when the classification was NC and the value of the specific parameter was identified in the C group.

 

Accuracy, , is the percentage of positive and negative predictions that were correct and responded to the question: “how many classrooms have been correctly labeled out of all the 29 classrooms?”. Precision, , is the percentage of positive predictions that were correct and responded to the following question: “how many of those classrooms that have been labeled as compliant classrooms, i.e., C, actually respect the threshold?”.

 

3. RESULTS

 

3.1. Measurement Results

 

Table 1 shows the results of the acoustic measurements in the primary school classrooms. The measured acoustical parameter range covers a wide span of values for all the considered parameters and represents most of the classroom environment typologies[6].

 

3.2. Relationship Between the Acoustical Parameters

 

The two-tailed significant correlations between the acoustical parameters measured in the classrooms corroborate the outcomes that had already been found in Astolfi et al. (2019) [1]. In particular, it was found:

• a strong negative relationship between the reverberation time under occupied conditions T20, and the C50 and U50 speech intelligibility indexes, which seems to suggest that it is necessary to use only one quantity to characterize classroom acoustics;
• a positive and significant correlation between T20 and T20_{_e}, which suggests the necessity of only characterizing classrooms under empty conditions;
• a very close connection between the central and mean values of both the C50 and U50 quantities, which indicates that only one measurement in the center of the room need to represent the behavior of the whole classroom, in terms of speech intelligibility;

 

Table 1: The acoustic parameters measured for each classroom. Standard deviations are indicated in brackets, when available; n.a. stands for not available. It is also indicated whether the classrooms belonged to the compliant group, i.e., C, or to the non-compliant group, i.e., NC, in the case of groupings (a) and (b). The values that do not comply with the threshold, i.e., False Positive (FP) and False Negative (FN), based on the different groups, are highlighted in bold when considering (a), in italics when considering (b), and in bold italics if the value does not comply with the (a) or (b) subdivisions.

 

Starting from the significant correlations, regressions analyses were carried out. High R² values were found for all the significant correlations, except for L_{S_ref} and T20, for which the relationship was weak. From correlations and regression analyses it is possible to assume that classroom acoustics can be fully characterized from just a single measure, e.g., T20 or C50_{_ctr},since it would be possible to estimate the other parameters through robust equations. C50 in the central position in occupied settings, i.e., C50_{_ctr}, can be considered as one of the most effective quantities for measurements inside a class room to investigate classroom acoustics. It is preferable to T20 and U50, because of its easier measurement procedure.

 

As far as the relationship between T20 and T20_{_e} is concerned, a close correlation emerged from the regression analyses, which suggests that only the empty condition is needed to be considered for an experimental survey.

 

3.2. Compliant and Non-compliant Classrooms

 

In order to categorize classrooms compliant (C) or non-compliant (NC), the classrooms were divided into two groups twice, i.e., assuming arbitrary T20 cut-off values of 0.8 s (a) and of 0.6 s (b), respectively. The objective was to ascertain the threshold that divides C from NC for each acoustical parameter in both of these cases.

 

The T20 cut-off value of 0.8 s was chosen as being representative of moderate requirements for class room acoustics since: (i) it is toward the upper range of admittable reverberation times in comfortable classrooms used for learning [7, 8], (ii) it is the optimal value advised in the case of the refurbishment of primary school classrooms and it is recommended for older pupils [9], and (iii) it is also recommended for speaking to older pupils [10]. The same T20 threshold of 0.8 s was used in Astolfi et al. (2019) [1] to discriminate between good acoustics and bad acoustics classrooms. In [1], the pupils’ perception of noise disturbance and their sense of fitting in at school differed significantly for the two groups, and the average subjective score was higher in bad acoustics than in good acoustics for the former and lower for the latter, respectively.

 

On the other hand, a T20 value equal to 0.6 s is the cut-off value that is required in the case of more severe classroom acoustics requirements. It was chosen since: (i) it is recommended in the most recent standards on classroom acoustics, such as the DIN 18041:2016 [11] and the UNI 11532-2:2020 standards [12], (ii) there is evidence that 0.6 s is the optimal value to guarantee a better learning perfor mance for primary school students [13]. A reverberation time (Tsoll) of 0.6±0.1 s, from 0.250 kHz to 4 kHz, is considered optimal, according to the DIN 18041:2016 [11] and UNI 11532-2:2020 standards [12], for rooms dedicated to communication with the simultaneous presence of several people speaking in the classroom, as is the case during primary school lessons.

 

Table 2 reports the descriptive statistics of grouping (a) together with the information if the two groups differ significantly, i.e., if p-value is below 0.05.

 

As can be seen from Table 2, there is no significant difference between the C and NC groups for the L_{N_sil}, L_{S_ref} or mL_S parameters, since the results of the Mann–Whitney U Test (MWU) are greater than 0.05.

 

The same procedure was done for grouping (b) that was made by imposing T20 equal to or below 0.6 s and then higher than 0.6 s to attribute classrooms to the C and NC groups, respectively. Once the classrooms had been assigned to the NC or C group on the basis of the T20 criterion, the descriptive statistics were calculated for each parameter as shown in Table 3.

 

Table 4 shows the AUC and the threshold values of the acoustical parameters obtained for the two groupings (a) and (b), which were used to subdivide the classrooms into C and NC. By comparing the thresholds in Table 4 with the optimal values of the acoustical parameters reported in literatureit is possible to label the two groups of classrooms as compliant (C) or non-compliant (NC), respectively, considering the upper limit of the admissible values in the case of T20 for C. The close correlation between the parameters determines the pairing of the thresholds for the two groups with the optimal values.

 

Table 4 also reports the accuracy and precision of the thresholds identified through the MSqD from the ROC curve for each acoustical parameter in labeling the classrooms in the two categories. Table 1 shows the values that do not comply with the threshold of the specific parameters, i.e., False Positive (FP) and False Negative (FN), which are in bold, in italics or in bold italics, on the basis of a group belonging to (a) or (b).

 

Table 2: Descriptive statistics of the acoustical parameters with a subdivision into C and NC for grouping (a). Standard deviations are indicated in brackets, while n.a. stands for “not available”. The p values that were obtained with the MWU test are also shown.

 

 

Table 3: Descriptive statistics of the acoustical parameters with a subdivision into C and NC for grouping (b). Standard deviations are indicated in brackets, while n.a. stands for “not available”. The p values that were obtained with the MWU test are also shown.

 

 

In short, Table 4 shows, in bold, the thresholds of the parameters that can be used alternatively to characterize classroom acoustics, according to the different level of performance required in the class room, that is, moderate performance, as with grouping (a), or severe performance, as with grouping (b). In the case of moderate requirements, such as for the refurbishment of existing classrooms or in the case of older pupils [9], as well as in the case of specific requirements for voice support [10], where a slightly higher reverberation time than 0.6 s is required, thresholds (a) are preferable. In this case, the parameters that are preferable to use alternatively for classroom acoustic characterization are T20, T20_{_e}, C50_{_M}, C50_{_ctr} and U50_{_M}. In the case of a severe performance requirement, that is, in the case of the respect of standard requirements, thresholds (b) are advised. T20, C50_{_M} and C50_{_ctr} result to be the most accurate and precise parameters for an alternative use to fully characterize class room acoustics for both moderate and severe requirements.

 

Table 4 also shows the paucity of cases belonging to the C group for the second grouping (b). A similar amount of data is needed for both groups to improve the statistical analysis and to have more robust results, like those in the case of grouping (a).This will be one of the aims of future research and it will be achieved by boosting the sharing of acoustical measures across Europe, since Italian schools are generally hosted in typical Southern European buildings, which have high ceilings and plaster walls, and are thus associated with higher reverberation time values.

 

Table 4: Area under the curve (AUC), thresholds identified through the minimum squared distance from the ROC curve and their accuracy and precision for groupings (a) and (b). The parameter data that can be used alternatively to characterize classroom acoustics, according to the different levels of performance required in the classroom, are shown in bold.

 

 

4. CONCLUSIONS

 

The obtained results show that most of the usually considered parameters are closely correlated. Reverberation time, T20, or speech clarity in the central point, C50_{_ctr}, or average across measurement positions, C50_{_M}, can all be used as the most representative parameters to characterize classroom acoustics. Therefore, using either T20 or C50, it is possible to estimate the useful-to-detrimental ratio, U50, which is the parameter that is most closely related with speech intelligibility, as it accounts for both noise and room acoustic defects.

 

In order to reduce the measurement points to a minimum number, it is advisable to first characterize classrooms by means of the speech clarity parameter in the central position, C50_{_ctr}.

 

Since the reverberation in occupied and unoccupied conditions resulted to be positively and significantly correlated, it is possible to state that it is sufficient to measure T20 in unoccupied conditions, using a procedure that gradually becomes smoother and faster, and from which the other related parameters can be estimated.

 

New thresholds for classroom acoustics parameters which discriminate the acoustic quality of primary school classrooms between compliant and non-compliant are provided for moderate and severe requirements. For moderate requirements the thresholds are 0.8 s for T20, 0.9 s for T20_{_e}, 67 dB(A) for LN_{_gr}, 3 dB for C50_{_M} and C50_{_ctr}, 1 dB for U50_{_M} and U50_{_ctr.} For severe requirements the thresholds are 0.6 s for T20, 0.9 s for T20_{_e}, 5 dB for C50_{_M} and 6 dB for C50_{_ctr}. Moderate limits are intended for older pupils and for refurbishment, whereas severe limits are intended for younger pupils and in the case of new schools. Both limits also support the teacher’s voice in the classrooms. However, the new thresholds are not intended to be definitive since they are not based on studies testing performance or subjective perception of pupils.

 

5. REFERENCES

 

1. Astolfi A., Puglisi G. E., Murgia S., Minelli G., Pellerey F., Prato A., and Sacco T. (2019), The influence of classroom acoustics on noise disturbance and well-being for first graders, Frontiers in Psychology 10, 1-20.

2. UK Department of Transport (2012). Available at: https://www.gov.uk/ government/publications/guid ance-on-road-classification-and-the-primaryroute-network (accessed October 11, 2021).

3. Croux C., and Dehon C. (2010), Influence functions of the Spearman and Kendall correlation measures, Journal of the Italian Statistical Society, 19, 2010-40.

4. Hajian-Tilaki K. (2018), The choice of methods in determining the optimal cut-off value for quantitative diagnostic test evaluation, Stat. Methods Med. Res. 27 (8), 2374–2383.

5. Zou H.K., O’Malley A.J., Mauri L. (2007), Receiver-Operating Characteristic Analysis for Evaluating Diagnostic Tests and Predictive Models, Circulation 115(5), 654-657, 2007.

6. Secchi S., Astolfi A., Calosso G., Casini D., Cellai G., Scamoni F., Scrosati C., and Shtrepi L. (2017), Effect of outdoor noise and façade sound insulation on indoor acoustic environment of Italian schools, Applied Acoustics, 126, 120-130.

7. Yang W., and Bradley J.S. (2009), Effects of room acoustics on the intelligibility of speech in classrooms for young children, J. Acoust. Soc. Am. 125(2), 922–932.

8. Hodgson M., and Nosal E.M. (2002), Effect of noise and occupancy on optimal reverberation times for speech intelligibility in classrooms, J. Acoust. Soc. Am. 111(2).

9. Building Bulletin 93 (2015), Acoustic design of schools: performance standards, Department for Education, London.

10. Calosso G., Puglisi G. E., Astolfi A., Castellana A., Carullo A., and Pellerey F. (2017), A one-school year longitudinal study of secondary school teachers’ voice parameters and the influence of classroom acoustics. J. Acoust. Soc. Am., 142(2), 1055-1066.

11. DIN 18041, 2004-05 (2016), Hörsamkeit in kleinen bis mittelgroßen Räumen (Acoustical quality in small to medium-sized rooms), Deutsche Institut für Normung, Berlin.

12. UNI 11532-2 (2020), Caratteristiche acustiche interne di ambienti confinati - Metodi di progettazione e tecniche di valutazione - Parte 2: Settore scolastico / Acoustic characteristics of indoor environments – Design methods and evaluation techniques – Part 2: school sector, Ente Italiano di Normazione.

13. Minelli G, Puglisi G.E., and Astolfi A. (2021), Acoustical parameters for learning in classroom: a review, Building and Environment (in press).

 

---

¹ arianna.astolfi@polito.it

² greta.minelli@polito.it

³ giuseppina.puglisi@polito.it