Globally convergent blind equalization algorithms for complex data systems

A set of memoryless blind adaptive equalization algorithms for nonminimum phase complex data systems is proposed and evaluated on the basis of admissibility. The algorithms are based on variations of a minimax cost on the equalizer output and actually take the form of gradient descent of linearly constrained convex cost functions. These investigations represent a systematic study based on nontrivial generalizations of admissible designs developed for real data systems. It is shown that for one such candidate generalization admissibility holds.<<ETX>>