Blind multivariable equalization

This paper gives an overview of several algorithms which have been proposed in the context of blind (i.e. without training sequence) multivariable (i.e. oversampled or multichannel) equalization. We focus first on the compatibility of previously known blind equalization (BE) algorithms with this relatively new framework. Then, two basic approaches are presented, explicitly dealing with the multivariable context. These two approaches are the subspace and the linear prediction methods. Finally, we emphasize on the validity of the underlying hypotheses (absence of common zeroes, given degree of the channel) in practical situations. These considerations serve as guidelines for discussing further research on blind multivariable equalization.