Blind channel identification and signal recovery by confining a component of the observations into a convex-hull of minimum volume

Abstract

This paper addresses the problems of the blind channel identification and signal extraction in a linear mixture of bounded complex sources. We present a blind criterion that solves these two related problems by confining a linear component of the observations into a convex-hull of minimum volume. The proposed criterion has its minima only at identification of the subspace of one of the unmixed components of the observations, allowing, therefore, a robust channel identification and signal recovery.