A Weighted Soft-Max PNLMS Algorithm for Sparse System Identification

Abstract

This paper presents a new Proportionate Normalized Least Mean Square (PNLMS) adaptive algorithm using a soft maximum operator for sparse system identification. To provide a high rate of convergence, soft maximum operator is employed along with a weighting factor, which is proportional to an estimation of output mean square error (MSE). Simulation results show the superiority of the proposed algorithm over its PNLMS-based counterparts.

Keywords

• sparse adaptive filter
• system identification
• echo cancelation
• soft maximum