An Enhanced IAF-PNLMS Adaptive Algorithm for Sparse Impulse Response Identification

Abstract

This correspondence presents an individual-activation-factor proportionate normalized least-mean-square (IAF-PNLMS) algorithm that (during the adaptive process) uses a new gain distribution strategy for updating the filter coefficients. This strategy consists of increasing the gain assigned to the inactive coefficients as the active ones approach convergence. For such, whenever a predefined threshold is crossed during the learning process, a new gain distribution is carried out, rather than to assign gains proportional to coefficient magnitudes as the IAF-PNLMS algorithm does. This new version of the IAF-PNLMS algorithm leads to a better distribution of the adaptation energy over the whole learning process. As a consequence, for impulse responses exhibiting high sparseness, the proposed algorithm achieves faster convergence, outperforming the IAF-PNLMS and other well-known PNLMS-type algorithms.