Real time acoustic echo cancellation technique based on normalized least mean square method

In this paper, we study real time acoustic echo cancellation technique based on NLMS (normalized least mean square) method and propose a method to improve conversation quality by the enhancement of acoustic echo cancellation technique performance during DT (double talk) section. In the environment that echo path changes slowly, proposed method stores past converged adaptive filter coefficients periodically and uses them when the DT section is started, instead of using complex DTD (double talk detector) algorithm to obtain the precise starting point of DT. Proposed method gives just small additional calculation and shows performance enhancement in terms of the remained echo during DT section.