Robust Nonlinear Echo Canceller with Adaptive Volterra Filter Applying VAD and Echo-Path Delay Estimation

M.A. Tinati, T.Y. Rezaii, and A. Rastegarnia

Keywords

Nonlinear echo cancellation, Volterra filter, Voice activity detection

Abstract

In this paper, a fast reliable and computationally efficient algorithm is proposed to solve the nonlinearity problem in line echo cancellation. Adaptive volterra filters, AVFs, are used as the basic tools to deal with such nonlinearities. We have made modified AVFs to make them robust and flexible, so they are more applicable to various problems like variable echo-path delays, variable nonlinearity intensities, high power echo component cancellations, speech signals with sparse voiced frames (slow speaker), high complexities, reliability, etc. In this new structure, an utmost attempt has been made to consider all of the practical problems while the minimum simplifying assumptions are made to achieve the best performance. The performance comparison of this structure with the other techniques in the same conditions and assumptions is performed and the results are the evidence of robustness and computational efficiency of the proposed structure compared with the others.

Important Links:



Go Back