Blind Source Separation Based on Genetic Algorithm

Recovering the unobserved source signals from their mixtures is a typical problem in array processing and data analysis. Independent component analysis (ICA) is a new and powerful method to solve this problem. JADE (joint approximate decomposition of eigen matrices) based on 4th-order cumulants is a common approach for ICA. However, it can only get approximate solution when k＞2 and the results are not accurate. In this paper, a blind source separation based on genetic algorithm is proposed. The analysis and simulations suggest that the scheme is feasible and effective.