On length adaptation for the least mean square adaptive filters

We address the problem of length adaptation for adaptive FIR filters implementing the LMS algorithm. In a previous paper, the authors have proposed an algorithm to adjust the length of the adaptive filter toward the optimum for uncorrelated inputs. Here, we extend the previous work to the case of correlated input sequences with emphasis in echo cancellation framework. Analytical expressions of the mean squared error and optimum filter coefficients are presented from which a new method of length adaptation is derived. Simulations, comparing our proposed algorithm with other related approach, in echo cancellation framework, are presented.