Blind multichannel adaptive MMSE equalization with controlled delay

The article addresses the problem of optimal blind linear symbol recovery, using the channel diversity induced by a sensor array or time oversampling. We present a technique allowing the computation of a minimum mean-square error (MMSE) equalizer, based on the optimization of quadratic second-order functions. The proposed technique improves on existing multichannel equalization methods, in that previous methods generally build on criteria which are optimal in the sole noise free context. Our criterion also allows a free choice of the delay for symbol recovery. As a consequence, the MMSE equalization performance can be enhanced through the use of an optimal delay. To this end, a performance analysis is conducted in order to investigate some of the links between the delay and the symbol estimation accuracy