An improved partial Haar dual adaptive filter for rapid identification of a sparse echo channel

Recently, a coupled echo canceller was proposed that uses two short adaptive filters for sparse echo cancellation. The first filter operates in the partial Haar domain and is used to locate the channel's dispersive region; the second filter is then centered around this location to cancel the echo in the time domain. In this paper, we propose feasible solutions to improve the performance of this partial Haar dual adaptive filter (PHDAF) in practical applications. These include: (1) alleviating the dependence of the PHDAFs performance on the echo-path impulse response's bulk delay; (2) improving the tracking performance of the PHDAF in response to abrupt changes in the echo path; and (3) extending the original PHDAF structure to support the cancellation of multiple echoes. The proposed algorithmic solutions exploit the Haar transform's polyphase representation and make use of a novel peak tendency estimator (PTE) based on Dezert–Smarandache theory (DSmT). The improved PHDAF is evaluated in terms of its mean-square error (MSE) curves and its mean time to properly locate a dispersive region for different SNRs. Results show that enhanced performance can be obtained using the proposed solutions at a minimal increase in computational cost.