Reformulation of linearly constrained adaptation and its application to blind equalization

This work introduces a reformulation of general linearly-constrained equalization using a fixed prefilter plus adaptive equalizer subject to a single constrained tap. This connection is exploited to simplify and improve an early result of Lucky (1965) on equalization and to derive an optimum, though nonblind, linear constraint for a broad class of cost functions. Based on the insights gained, we briefly suggest that a globally convergent adaptation scheme can be formulated by either generalizing the parameter constraint or generalizing the specific cost to a general convex cost and also note that such a scheme may be applied to blind equalization.<
>