Acoustic echo and noise canceler improvements for hands-free telephones

We propose a new strategy for echo cancellation, where we switch between the least mean squares (LMS) and subband NLMS (SNLMS) algorithms, to achieve fast convergence and high steady state echo return loss (ERL). We start with a low order NLMS filter to achieve fast convergence. Later, we switch to a high order SNLMS filter to achieve high steady state ERL. In order to combat environmental noise, we propose cascading an adaptive noise canceling filter. The noise canceling filter converges in the mean to the optimal causal Wiener solution to filter the noise from the near-end signal. Finally, simulation results show the performance of the proposed echo and noise cancellation strategies.