Blind equalization via linearly constrained minimum variance processing

Abstract

A new cost function and a new adaptive structure for blind equalization of communications systems with FIR filters are proposed. It is first shown that the cost function of a previously proposed blind equalization algorithm can be expressed in a similar manner to that of the linearly constrained minimum variance (LCMV) problem (which arises in array processing). This new viewpoint permits a new understanding of the convergence behavior of the previously published technique, as well as the development of new approaches to blind equalization. In particular, a new "RLS-like" algorithm is developed that exhibits a convergence rate much faster than previously published algorithms of its class, with a modest increase in computational complexity.