Novel algorithm for independent component analysis with flexible score functions

Independent component analysis (ICA) refers to the recovery of a set of independent sources when only the mixtures of these sources with unknown coefficients are observed. It is a mainstream technique for blind source separation (BSS). This paper introduces a method for blind source separation without any knowledge of their probability distributions. This is achieved under a maximum likelihood framework by considering the parametric density mixture model and Pearson system model. As a result, a novel explicit ICA algorithm with flexible score functions to various marginal densities is obtained. Simulation result shows that the proposed algorithm is able to separate a wide range of source signals, including sub-Gaussian and super-Glaussian sources, symmetric and asymmetric sources.