Normalized LMS algorithm with evolutionary change of orthogonal correction factors

This paper proposes a normalized LMS algorithm which automatically determines the number of orthogonal correction factors using an evolutionary method. The evolutionary method increases or decreases the number of orthogonal correction factors by comparing the squared output errors with some threshold. The experimental results show that the proposed algorithm has not only a fast convergence rate and a small steady-state estimation error but also lower computational complexity than the existing algorithms with multiple input vectors.