A two-stage Bayesian approach for complex blind channel equalization

The paper extends a class of optimum Bussgang equalizers (based on Bayesian concepts) to any non-separable bidimensional modulation format, such as general PSK. The inherent flexibility of the considered operator with respect to known methods (such as constant modulus) allows its behavior to be optimised with regard to ambiguity and accuracy problems. The use of this operator in two stages (coarse plus fine scheme) allows easy convergence control for severely distorted channels