Blind Equalization Based On Blind Separation with Toeplitz Constraint

Blind equalization (BE) has been modeled as a blind source separation (BSS) problem and achieved using BSS algorithms. One drawback of on-line algorithms considered in previous work for blind equalization is their slow convergence. We show that the Toeplitz structure of the mixing matrix and the separating matrix in the BSS model for BE can be exploited to obtain faster convergence and better performance. In addition, a constraint on the length of the equalizer impulse response can provide further improvement. For sources with independent in-phase and quadrature parts, we can use I/Q constrained BSS technique to do equalization and achieve phase recovery. We use the well-known equivariant adaptive source separation via independence (EASI) algorithm to illustrate the ideas, although the approach we describe is more generally applicable.