Adaptive combination with improved performance for sparse system

Abstract

We propose an adaptive combination of a proportionate normalized least-mean-square with individual activation factors (IAF-PNLMS) and a normalized mean-square (NLMS) for the sparse system. The IAF-PNLMS has the fastest initial convergence rate among the algorithms for the sparse system. The NLMS has a low misalignment for various systems. To obtain both fast convergence rate and a low misalignment, we derive the proposed algorithm through adaptive combination algorithm of the IAF-PNLMS and the NLMS. We simulate to show the proposed algorithm has better performance than the conventional algorithms for the sparse system in terms of convergence rate and steady-state performance.