Pseudoanechoic blind source separation with improved Wiener postfilter ∗

The pseudoanechoic model was proposed recently to simplify the parameter estimation in blind source separation based on frequency-domain independent component analysis. In the method, after separation based in the pseudoanechoic model a time-frequency Wiener postfilter to improve the separation is applied. In this contribution, a deeper analysis of the working principles of the Wiener postfilter is presented. Furthermore, a variation of this postfilter to improve the performance using the information of previous frames is introduced. The improvements obtained through the new method are evaluated in an automatic speech recognition task and with the PESQ objective quality measure. The results show an increased robustness and stability of the proposed method. Keywords— Pseudoanechoic model, Blind source separation, Automatic speech recognition, Wiener postfilter