Prediction of the Effect of a Sound-Masking System

Supervisor: Murray Hodgson
Researcher: Yizhoung Lei

Introduction

So far, speech privacy has been an important and hardly achievable issue for open-plan offices, especially in natural-ventilation buildings with low background-noise levels. This research tries to estimate the effect of a sound masking system (SMS) on improving speech privacy in the CIRS Building Simulation Software (BSS) Lab, and how the Lombard effect influences it.

Methods

Acoustical measurements were performed in the BSS Lab in CIRS. Three acoustic parameters were measured and analyzed – Reverberation Time (RT), Background Noise Level (BNL), and Speech Intelligibility Index (SII).

To analyze the effect of a SMS and of the Lombard effect, the CATT-Acoustic software was used to predict Speech Transmission Index (STI) in the BSS Lab. A SMS was designed using the CATT-Acoustic software.

To predict talker voice levels considering the Lombard effect, an iterative model proposed by Hodgson, Steininger and Razavi*, was used.

*reference: M. Hodgson, G. Steininger, and Z. Razavi, “Measurement and prediction of speech and noise levels and the Lombard effect in eating establishments,” Acoustical Society of America, 121(4), 2023–2033 (2007).

Results

The acoustic environment in the CIRS BSS Lab was evaluated. The BNL, the RT and the SII all exceeded the acceptable criteria. Therefore, acoustic improvements are needed in the BSS Lab. The Lab 3D model of the BSS Lab was proved accurate and used in predictions for estimating the effect of the SMS and the Lombard effect.

The SMS increases the background noise level to the desired values (NC40). The uniformity of the acoustic environment with the SMS in operation is acceptable. The SMS helps achieve excellent speech privacy in the BSS Lab, preserving good speech intelligibility. When the Lombard effect is considered, the values of STI increase but are still in the range of “excellent speech privacy”. The effect of SMS would be reduced, and sometimes canceled by the Lombard effect.

Conclusion

A prediction model has been created for the estimation of the acoustical parameters. The effect of a sound masking system on indoor speech privacy was computed and evaluated. In all situations discussed in this report, the speech privacy became poorer due to the Lombard effect. Without considering the Lombard effect, the improvement on speech privacy, caused by the SMS, could be easily overestimated.

Read Full Text

Prediction of the Effect of a Sound-Masking System in CIRS Including the Lombard Effect: Yizhong Lei