Blind separation of sources by differentiating the output cumulants and using Newton’s method

Abstract

This paper addresses the Blind Separation of non-Gaussian Sources for instantaneous mixtures. It is proven that the estimation of the separating system can be based on the cancellation of some second partial derivatives of the output cross-cumulants. These equations can be seen as a subset of those solved by JADE. Finally, we present an algorithm which is based on Newton’s method and exhibits a fast rate of convergence.