Mean square error analyses of adaptive blind source separation algorithms

Although several simple and useful adaptive blind source separation algorithms have been developed in the scientific literature, few analyses of their second-order statistical properties have been derived. In this paper, we give the complete details of a procedure for determining the average steady-state mean square error of a blind source separation algorithm. We then apply the procedure to nine existing blind source separation algorithms employing fourth-order separation criteria. Simulation results verify the accuracy of the analysis method.