Unbiased Blind Adaptive Channel Identification and Equalization

The blind and adaptive equalization or identification of communication channels is a problem of important theoretical and practical concerns. Many recently proposed solutions typically exploit the diversity induced by sensor arrays or time oversampling, leading to the so-called second-order multichannel techniques. However, some of these methods build on conditions or equations which are valid in the noise free case only. This notably includes the Prediction Error (PE) method which, tough very attractive, lacks optimality in the context of noisy data in the sense that it naturally yields biased estimates of the channel coefficients. In the adaptive context, we investigate solutions to this problem by exploiting the rank-deficiency of the multichannel covariance matrices. We derive a low-cost algorithm for the adaptive computation of unbiased linear predictors. This algorithm does not require the knowledge of the noise variance and can be used to improve the channel estimates provided b..