A unified approach for sparsity-aware and maximum correntropy adaptive filters

Abstract

Adaptive filters that employ sparse constraints or maximum correntropy criterion (MCC) have been derived from stochastic gradient techniques. This paper provides a deterministic optimization framework which unifies the derivation of such algorithms. The proposed framework has also the ability of providing geometric insights about the adaptive filter updating. New algorithms that exploit both impulse responses sparsity and MCC are proposed, and an estimate of their steady-state MSE is advanced. Simulations show the advantages of the proposed algorithms in the identification of a sparse system with non-Gaussian additive noise.