A new kind of on-line prediction BSE method based on genetic algorithm

The blind signal extraction (BSE) approach based on an on-line predictor is put forward and developed recently. All the on-line predictor extraction algorithms are optimizing process based on gradient algorithm at present. The choice of initial values and learning rates could influence the performance of the algorithm. A kind of BSE approach based on genetic algorithm is put forward in the paper. The influence of initial values and learning rates on the performance of the algorithm is avoided in the kind of BSE approach based on genetic algorithm. Otherwise, various source signals are extracted by changing the order of the predictor without performing the deflation process in the algorithm. The precision is improved because the error of the extraction signals forward is not cumulated for the extraction signals backward. Simulation results illustrate the efficiency and better performance of the algorithm by separately simulating the on-line predictor extraction algorithms based on gradient algorithm and the algorithm put forward in the paper.