In this paper We propose a neural network algorithm for independent component analysis(ICA) which can separate of right-skewed and a fat right-hand tail source signals with self-adaptive activation functions . The ICA algorithm in the framework of fast converge Newton type algorithm, is derived using the parameterized generalized K-distribution density model. The nonlinear activation function in ICA algorithm is self-adaptive and is controlled by the shape parameter of generalized K-distribution density model. To estimate the parameters of such activation function we use an efficient method based on maximum likelihood (ML). Computer simulation results confirm the validity and high performance of the proposed algorithm .
展开▼
