Blind Source Separation Using Quadratic form Innovation

Abstract

Blind source separation (BSS) is an increasingly popular data analysis technique with many applications. Several methods for BSS using the statistical properties of original sources have been proposed, for a famous one, such as non-Gaussianity, which leads to independent component analysis (ICA). This paper proposes a blind source separation method based on a novel statistical property: the quadratic form innovation of original sources, which includes linear predictability and energy (square) predictability as special cases. A gradient learning algorithm is presented by minimizing a loss function of the quadratic form innovation. Also, we give the stability analysis of the proposed BSS algorithm. Simulations verify the efficient implementation of the proposed method.