An Algorithm Based on Nonlinear PCA and Regulation for Blind Source Separation of Convolutive Mixtures.

This paper proposes a method of blind separation which extracts independent signals from their convolutive mixtures. The function is acquired by modifying a network's parameters so that a cost function takes the minimum at anytime. Firstly we propose a regulation of a nonlinear principle component analysis (PCA) cost function for blind source separation of convolutive mixtures. Then by minimizing the cost function a new recursive least-squares (RLS) algorithm is developed in time domain, and we proposed two update equations for recursively computing the regularized factor. This algorithm has two stages: one is pre-whitening, the other is RLS iteration. Simulations show that our algorithm can successfully separate convolutive mixtures and has fast convergence rate. [ABSTRACT FROM AUTHOR]