Unimodal blind adaptive channel equalization: an RLS implementation of the mutually referenced equalizers

The blind, direct, and adaptive linear equalization of possibly rapidly time varying communication channels is a problem of important theoretical and practical concerns. In this paper a new solution is developed for the equalization of oversampled channels (temporally or spatially) which improves on most existing related contributions. First, the proposed criterion is purely quadratic, hence it ensures global convergence. Then, the algorithm provides a complete set of channel inverses, resulting in robustness properties w.r.t. reconstruction delay problems. Finally, and more importantly, all standard adaptive filtering structures can be used to implement the method. Among these, the RLS-based solution reaches convergence after a hundred iterations only, under certain conditions.