Abstract: A modified hierarchical least mean square (HLMS) algorithm is proposed. The response of each sub-filter in the hierarchical level is modified to improve the performance of the HLMS algorithm, making the filtering for each sub-filter possible. Simulation results indicate that the modified HLMS algorithm not only maintains fast convergence, but also provides improvement of the mean square error of the impulse response by one order of magnitude with respect to the original HLMS algorithm under the same conditions.

Key words: hierarchical filters, least mean square algorithm
, FIR transversal filtering