A novel target speaker dependent postfiltering approach for multichannel speech enhancement

In this article, we present a target speaker dependent speech enhancement system, to enhance a specific target talker in presence of real life background noises. The proposed system uses a multi-channel processing stage to produce a noise reference signal. This noise reference signal is further used, to not only compute the residual noise statistics from the multichannel stage output, but also to learn the noise subspace for a Non-negative Matrix Factorization (NMF) based post filtering. In our evaluations using GRID database [1], both speaker dependent and speaker independent state-of-the-art enhancers have been considered. Our proposed approach not only outperforms the speaker independent approaches significantly, but also shows improvement over a recently proposed speaker dependent approach. Finally, an online version of the algorithm has also been proposed, with a slight compromise in the performance compared to the batch system.