Candidate admissible blind equalization algorithms for QAM communication systems

The authors propose and investigate the admissibility of a group of memoryless blind equalization algorithms for nonminimum phase quadrature amplitude modulation (QAM) systems that are based on stochastic gradient descent of linearly constrained convex cost functions. Two such algorithms for compleeata systems are introduced, one of which is shown to be admissible for square symbol constellations. This investigation also identifies some issues that arise out of moving from real to complex systems such as the design of parameter constraints that account for channel-induced phase rotation and the potential nonuniqueness of the optimal equalizer parameter setting. Other remarks are made about nonsquare QAM constellations and autoregressive moving average (ARMA) equalizer parametrizations