Fuzzy neural-based learning rate adjustment for gradient based blind source separation

Independent component analysis (ICA) algorithms have been proposed to solve blind source separation (BSS) problem in recent years. T he gradient algorithm is a popular method deals with separating independent signal step by step with learning rate. In this paper, consider to balance the mis-adjustment and the speed of convergence, the leaning rate will be computed in fuzzy neural network (FNN) depended on the second-order and higher order correlation coefficients of output components of BSS. To enhance the performance of the FNN-based learning rate, the FNN is optimization by particle swarm optimization algorithm. Finally, simulation results are shown to illustrate the effectiveness of the proposed method.