Quasi-Blind Source Separation Algorithm for Convolutive Mixture of Speech

Based on the assumption that there are short periods of time in which only one source is active, a new approach for convolutive blind source separation (quasi-BSS) is proposed, which does not require the signals to be independent or identically distributed. In order to obtain a low-complexity iterative solution to the separation filters, an optimization method based on the affine projection adaptation algorithm (APA) with non-orthogonal constraint is used. The effect of noise on the convergence of the algorithm is given theoretically. Simulations based on synthetic data and real room recordings show excellent noise immunity, a high convergence rate, and good tracking capability of the proposed method