Direct blind equalization with best delay by channel output whitening

In this paper, we describe methods for computing fractionally spaced blind equalizers directly from second-order statistics of the observations. The equalizers with all possible delays are computed simultaneously and the one with the best delay is picked by a blind constant modulus index. Our algorithm utilizes the equalizer output whitening property of the blind equalizers, i.e., output whitening yields equalization; thus, no channel identification is required. This may reduce the performance degradation that is caused by channel estimation errors of channel estimation based approaches such as the subspace methods. In addition, existing channel output whitening-based approaches require nonlinear optimization and are unable to choose equalizer delays. In contrast, our algorithm has simple closed-form solutions for equalizers with all possible delays. Comparison of our algorithm with some other algorithms in the similar category is given, which shows that our algorithm has more efficient computations. Simulations demonstrate the good performance of our method