Blind Source Separation Method Based on Real Coded Genetic Algorithm

In this paper, a new cost function based on the diagonalization of correlation matrices is proposed to perform blind source separation. This cost function can restrain cross-correlation of separated signals and be applicable to separate instantaneous or convolutive mixture of stationary or non-stationary signals. A real coded genetic algorithm is proposed to search the optimum solution. Computer simulation results demonstrate this algorithm has not only fast convergence performance but also high accuracy.