Estimation of the noise power in the NPVSS-NLMS algorithm

The normalized least-mean-square algorithm (NLMS) is frequently used in the context of acoustic echo cancellation (AEC). However, in order to meet the conflicting requirements of fast convergence and tracking versus low misadjustment, the variable step-size NLMS (VSS-NLMS) algorithms were found to be more suitable in practice. In this paper, we analyze the nonparametric VSS-NLMS (NPVSS-NLMS) algorithm that uses the power estimate of the background noise (i.e., the noise that corrupts the near-end speech) in order to control its step-size parameter. The estimation method of the noise power is based on a time-recursive averaging algorithm. Simulation results indicate the good performance of the NPVSS-NLMS algorithm.