Step Size Control of Shared-error Normalized Least Mean Square Algorithm for Acoustic Echo and Noise Canceller

We propose a step size control method for an acoustic echo and noise canceller (AENC) based on a sharederror normalized least mean square (SENLMS) algorithm. The conventional AENC based on the SENLMS algorithm can optimize both the acoustic echo canceller (AEC) and the adaptive noise canceller (ANC). However, it is difficult to set the step size parameters to prevent the divergence of the conventional AENC because of the optimization of both AEC and ANC. To solve this problem, we investigate the constraint for the step size parameters of the AENC with the SENLMS algorithm and propose a step size control method. The constraint for the step size parameters is derived from the mean square value of the a posteriori error. Then, the step size parameter of the AEC is determined by the nonparametric variable step size (VSS) NLMS algorithm. Finally, the step size parameter of the ANC is automatically determined by the constraint and the variable step size parameter of the AEC. Simulation results show that the proposed AENC with the VSS-SENLMS algorithm can simultaneously prevent the divergence and maintain its the capability to reduce background noise and acoustic echo.