---

Abstract:

Blind System Identification (BSI) is a fundamental problem which is known by different names: blind deconvolution, blind equalization, blind deconvolution or blind separation of convolutive mixtures. In all these instances, the objective is an inverse problem where, given an observed (multi-channel) system output, one seeks to identify the unknown channels and/or retrieve at best the unknown input signals of the considered system. BSI is often an ill-posed, difficult problem which is simplified in the context of communications system, by the addition of pilot sequences that are known at the receiver side. In such context, exploiting both pilot and data for the system identification is referred to as semi-blind processing. This tutorial is a brief review of these techniques and their potentials and limitations and is addressed to a wide audience with general background in signal processing