Speech reverberation suppression for time-varying environments using weighted prediction error method with time-varying autoregressive model

In this paper, a novel approach for the task of speech reverberation suppression in non-stationary (changing) acoustic environments is proposed. The suggested approach is based on the popular weighted prediction error (WPE) method, yet, instead of considering fixed reverberation prediction weights, our method takes into account the more generic time-varying autoregressive (TV-AR) model which allows dynamic estimation and updating for the prediction weights over time. We use an initial estimate of the prediction weights in order to optimally select the TV-AR model order and also to calculate the TV-AR coefficients. Next, by properly interpolating the calculated coefficients, we obtain the ultimate estimate of reverberation prediction weights. Performance evaluation of the proposed approach is shown not only for fixed acoustic rooms but also for environments where the source and/or sensors are moving. Our experiments reveal further reverberation suppression as well as higher quality in the enhanced speech samples in comparison with recent literature within the context of speech dereverberation