Channel Sparsity-Aware Diagonal Structure Volterra Filters for Nonlinear Acoustic Echo Cancellation

摘要

In this paper, we propose a channel sparsity-aware recursive least square (RLS) algorithm using sequential update for nonlinear acoustic echo cancellation (NAEC). The acoustic nonlinear echo path is modeled using the diagonal channel structure Volterra filters. The results are compared with the linear and functional-link artificial neural network (FLANN) filters. The improved performance is validated via computer simulations. The nonlinear echo canceller designed using a third-order Volterra filter offers the best echo cancellation performance while the canceller using the sparse third-order Volterra filter provides the compromised performance with a reduced computational load.