Blind equalization using cost function matched to the signal constellation

The constant modulus algorithm (CMA) has been proved to be very effective when used to equalize non-minimum-phase channels blindly. However, the CMA suffers from a residual intersymbol interference (ISI) and slow convergence when applied to high order non-constant modulus (e.g. QAM) constellations and the equalizer is an FIR filter. We propose a method that is able to provide a lower ISI and a faster convergence rate than the CMA, based on cost functions specifically matched to the signal constellation. It is different from other CMA methods, where the matching between a certain number of cumulants between the equalizer output and the signal constellation is the driving criterion, the cost function used in this paper tries to force the equalizer output to belong to the signal constellation. The superiority of the proposed approach with respect to conventional approaches is verified by simulation.