Adaptive Super-exponential Methods for Blind Multichannel Equalization

An adaptive super-exponential algorithm for multichannel fractionally-spaced blind equalization is proposed. In the noise-free case, oversampling the received signal in space or time leads to a rank-deficient covariance matrix of the corresponding vector process. This seems to be the major obstacle towards a multichannel adaptive super-exponential algorithm. In the absence of additive noise, the optimal multichannel equalizer setting can also be found as the solution of a suitably chosen quadratic cost function. In the presence of a moderate amount of noise, minimization of the same cost function should result in an equalizer setting close to the optimal one. The optimization, however, can now be performed recursively using a stabilized RLS algorithm. Experimental results with underwater acoustic communication data demonstrate the good performance of the suggested method.