A Robust Blind Sparse Source Separation Algorithm using Genetic Algorithm to Identify Mixing Matrix

T.-Y. Sun, C.-C. Liu, S.-T. Hsieh, S.-J. Tsai, and K.-Y. Li (Taiwan)


Blind source separation (BSS), sparse representation, genetic algorithm (GA), under-determined


In this paper, a novel identification of mixing matrix using genetic algorithm (GA) is proposed to deal with the blind sparse source separation (BSS) problem. A preprocessing filters the most of minor mixtures at first, and then represents the remainder in angle. Further, we regard a probable set of angle of mixing vectors as a chromosome of GA, and iterate the evolutionary loop to minimize the fitness function which summarizes the angle difference between mixtures and estimated mixing vector. In computer simulations, mixing matrixes with well condition and ill-condition are considered for testing, meantime several algorithms are carried them out also. It was demonstrated by simulation results that the proposed GA-based algorithm is superior in validation and effectualness than others.

Important Links:

Go Back