General block lms algorithm

In this paper, we analyze the conventional block-least-mean-square (BLMS) algorithm. Usual constraints such as real input data, steady-state analysis and positive adaptive step-size parameter are discarded. Some modifications are introduced in order that the new complex frequency-domain BLMS algorithm equals the former versions in case any of the constraints are imposed. Furthermore, if the steady-state analysis is considered, the proposed algorithm avoids the inversion of the autocorrelation matrix of the transformed input.