An on-line Blind Source Separation Algorithm for Temporally Correlated Signals

An on-line blind source separation algorithm is presented in this paper. By assuming that the sources are temporally correlated signals with white noises added in their measurements, blind source separation could be finished by using only second order statistics of partial observed samples in an iterative calculation mode. A special cost function is used to determine the eigenvector matrix of the observed signals´ covariance matrix, which doesn´t need the singular value decomposition (SVD) that commonly used in many methods. Simulations have been made to validate the effectiveness of the algorithm