An improved partial Haar dual adaptive filter for sparse echo cancellation

Abstract

This paper proposes the use of a peak tendency estimator (PTE) based on Dezert Smarandache Theory (DSmT) and fuzzy inference to overcome two inherent limitations of a recently proposed partial Haar dual adaptive filter (PHDAF) for sparse echo cancellation. These limitations include the dependence of the PHDAF's performance on the echo path impulse response's bulk delay as a result of the lack of shift-invariance of the wavelet transform, and the PHDAF's difficulty in quickly tracking a new dispersive region after an abrupt change in bulk delay occurs. The improved PHDAF is analyzed in terms of its mean-square error (MSE) curves as well as its mean time to properly locate a dispersive regions under different SNRs. The simulations show that enhanced performance can be obtained using the proposed solutions at a minimal increase in computational cost.