Approach to nonlinear blind source separation based on niche genetic algorithm

Blind source separation (BSS) is a class of methods that recover inaccessible independent original signals from unknown mixtures. This paper proposes the niche genetic algorithm in combination with nonlinear blind source separation to solve the global optimization of parameters. The mixing model is the well-known post-nonlinear (PNL) mixture. The natural gradient descent method is applied in minimizing mutual information to estimate the separation matrix. Niche genetic algorithm (NGA) is used to obtain the globally optimal coefficients of polynomials which estimate the inverse of nonlinear mixture function. The simulation is performed and waveforms of separated signals are approximately identical with source signals. Experimental results indicate that the proposed method of NGA can quickly and effectively get optimal resolution to nonlinear blind source separation. Compared to conventional approaches, the proposed method is characterized by high accuracy, fast convergence, and robustness against local minima.