A multistage approach for blind separation of convolutive speech mixtures

In this paper, we propose a novel algorithm for the separa-tion of convolutive speech mixtures using two-microphone recordings, based on the combination of independent com-ponent analysis (ICA) and ideal binary mask (IBM), together with a post-filtering process in the cepstral domain. Essen-tially, the proposed algorithm consists of three steps. First, a constrained convolutive ICA algorithm is applied to sepa-rate the source signals from two-microphone recordings. In the second step, we estimate the IBM by comparing the en-ergy of corresponding time-frequency (T-F) units from the separated sources obtained with the convolutive ICA algo-rithm. The last step is to reduce musical noise caused typi-cally by T-F masking using cepstral smoothing. The perfor-mance of the proposed approach is evaluated based on both reverberant mixtures generated using a simulated roommodel and real recordings. The proposed algorithm offers consider-ably higher efficiency, together with improved speech quality while producing similar separation performance as compared with a recent approach. Index Terms — Independent component analysis (ICA), ideal binary mask (IBM), estimated binary mask, cepstral smoothing, musical noise 1