Comparative study of new Blind Source Separation structures for two-channel Acoustic Noise Cancellation

This paper addresses the problem of Blind Source Separation (BSS) applied to Acoustic Noise Cancellation (ANC) schemes when two microphones are used for the sound pick-up. The proposed approach is based on an improved Forward BSS structure combined with a post-filter [1] in order to correct the inherent speech distortion brought by the Forward BSS structure. The performance of the proposed algorithm is compared to the performance of two Backward BSS structures, namely the classical Backward structure [2] and the adaptive solution proposed in [3]. Performances of the proposed algorithm are evaluated by the output SNR and the cepstral distance under various environments. Experimental results indicate that the proposed method outperforms the classical Backward structure and the adaptive one, especially in the critical case of closely spaced microphones.