A new ICA method based on chaos artificial fish swarm algorithm

Independent component analysis(ICA) is a blind source separation(BSS) technique, and also an optimization problem. The current ICA algorithms have the shortages of slow convergence and fall easily into the local optimum. A new method, which applied chaos artificial fish swarm algorithm(CAFSA) to ICA optimization calculation, called CAFSA_ICA, was proposed. Combined with the objective function based on negentropy maximum criterion, several artificial fishes(AFs) were initialized and parallel searched in the feasible domain of de-mixing matrix w. In this way, CAFSA_ICA achieved the fast global convergence. Meanwhile, ergodicity and pseudo-randomness of chaos searching were used to improve the convergence precision further, so as to make ICA get a better separation performance. It was tested on synthetic 4-channel signals and real EEG signals. SNR, PI and iteration time were evaluated. The results show that CAFSA_ICA own the satisfactory separation performance and the high separation efficiency, being worth researching ulteriorly.