A criterion for the separation of a linear-quadratic mixture of independent components

Blind identification of a linear mixture is a possible approach for source parameters estimation in array processing. The goal is to perform blind identification of a linear-quadratic mixture of independent and circular random components. The authors propose an approach based on the maximization of a criterion. They define a class of criteria appropriate to the linear-quadratic mixture. They show that the maxima of these criteria over the set of the unitary matrices allow to identify the mixture. They give the expression of such a criterion and an algorithm to maximise it. This problem is of interest in all areas involving second-order nonlinear mechanisms or systems in which the inputs are unobservable