Acoustic Echo Canceling in the Double-Talk Condition

In the echo canceling for hand-free set mobile radiotelephone or teleconference system, the double-talk is occurred when both the near-end and the far-end speakers talk simultaneously. The conventional adaptive digital FIR filter using algorithm such as LMS fails to track the echo path in this condition. Because the error signal is contaminated to the near-end signal to estimate the gradient correctly. The Correlation LMS (CLMS) and Extended Correlation LMS (ECLMS) algorithms have been introduced by authors to challenge the double-talk in the echo canceling system. In this paper we propose a new implementation of ECLMS algorithm in the wavelet domain called wavelet transform extended correlation LMS algorithm (WECLMS) to improve the speed of the convergence. The computer simulation results support the theoretical findings and verify the robustness of the proposed WECLMS algorithm in the double-talk situation.