Blind source separation based on moments matching

Abstract

The Blind Source Separation (BSS) is a fundamental and challenging problem in signal processing. A new method based on the fractional moments and simulated annealing is presented in this paper. Fractional moments are used as a new measure of separation criterion contrary to the traditional integer moments. This is inspired by the fact that fractional moments lead to an enhanced estimation of the probability density function (PDF). Simulated Annealing (SA) is selected as the optimization algorithm to avoid being trapped into local minima. Simulation results validate the applicability of the new strategy for BSS .