On the linearly constrained blind multichannel equalization

Working at baud rate regime (one sample per symbol), it is well known the equivalence between the linear prediction problem and a proper linearly constrained power minimization criterium [1]. However, the success of this criterium as a blind equalization technique is limited to the minimum phase channel condition. On the other hand, if it is assumed a multichannel model (several samples per symbol or several sensors), it has been shown, under certain hypotheses, the minimum phase character of the multivariate transfer function; this property is in fact which allows one of the main approaches of multichannel blind equalization as a multivariate linear prediction problem [2]. Our main goal is to introduce a family of adaptive algorithms dealing with the formulation of the blind equalization task as a linearly constrained cost function in order to generalize the baud rate case results: a detailed analysis of the mentioned cost functions is included and also supported by several computer simulations.