Blind estimation of speech transmission index and room acoustic parameters based on the extended model of room impulse response

Abstract

The speech transmission index (STI) and room-acoustic parameters, such as the reverberation time and clarity index, are essential to assess the quality of room acoustics. However, in everyday spaces occupied by people, it is difficult to obtain such parameters since the room impulse response (RIR) cannot be measured. Blind estimation of room acoustic parameters from observed signals without measuring RIR is therefore necessary. Although existing methods can estimate one of these parameters, a single parameter is inadequate to describe comprehensive subjective aspects. To this end, this paper proposes a method for estimating STI and five room-acoustic parameters simultaneously. The temporal amplitude envelope of a reverberant speech signal is mapped to the parameters of an RIR model for each sub-band. Instead of using Schroeder’s RIR model, the extended RIR model is used to approximate an unknown RIR so that the STI and room-acoustic parameters can be derived. We performed simulations to evaluate the proposed method in unseen reverberant environments. The root-mean-square errors between the ground-truths and estimated parameters suggest that the accuracy of the proposed scheme outperformed that with Schroeder’s RIR model and was close to the standard measurements.