A stable and robust ICA algorithm based on t-distribution and generalized Gaussian distribution models

We propose a novel independent component analysis (ICA) algorithm which enables one to separate mixtures of sub-Gaussian, super-Gaussian and Gaussian primary source signals. Alternative activation functions in the algorithm are derived by using parameterized t-distribution and generalized Gaussian distribution density models. The functions are self-adaptive based on estimating the high-order moments of extracted signals. Moreover, a stability condition of the proposed algorithm for separating the true solution is given. Simulation experiment results are presented to illustrate the effectiveness and performance of the proposed algorithm.