Step Adaptive Normalization Blind Source Separation Algorithm

Abstract

An algorithm of step adaptive normalization BSS(SAN-BSS) is proposed to solve the problem that the traditional switching BSS algorithms are sensitive to the types and the number of the source signals. The proposed algorithm improves the original ones’ stability by making use of the normalization mechanism to modify the cost functions, and realizes the adaptive updating of the step size by combining the signals’ separation process with the summation of the edge negentropy. The simulation results show that when the number of the source signals improves or the types of the signals change, the proposed algorithm can keep good separation effect. Compared with the original ones, the separation accuracy of our proposed algorithm improved 98%, and the number of iterations reduced nearly 60%, which improved the stability and the separation speed of the algorithm greatly.