Polyspectra-based, blind, MMSE, fractionally-spaced equalization of a cyclostationary signal

Abstract

Two weaknesses of polyspectra-based blind equalization are addressed. The first weakness involves a zero-forcing (ZF) characteristic, where intersymbol interference is eliminated with disregard for associated mean squared error (MSE). It is shown that a minimum-MSE (MMSE) constraint can be added for signals operating in colored Gaussian noise, with only a small increase in computational complexity and with performance enhancement under certain scenarios. The second weakness involves the constraint that the signal be stationary, which fails to exploit the cyclostationary features of most communication signals (W. A. Gardner). In particular, it is shown that polyspectra-based techniques can be used with a fractionally-spaced equalizer, with attendant performance boosts in timing insensitivity (S. H. Qureshi, 1985) and damaged-spectrum restoration (W. A. Gardner, 1991). The technique decomposes the signal into stationary streams, computing the individual and joint statistics of the partitions, and then solves the optimal Wiener equation.<>