Performance limitations of a new subband adaptive system for noise and echo reduction

Oversampled Subband Adaptive Filters (OS-SAFs) have shown good performance in echo and noise cancellation. In order to understand the ultimate limitations, two issues must be considered: convergence rate reduction, and asymptotic performance reduction resulting from aliasing and other effects. When the Normalized Least Mean Square algorithm is employed for OS-SAFs, the convergence rate is considerably decelerated due to the coloring of subband signals. We have already proposed new techniques for improving the convergence rate based on spectral emphasis and decimation of the subband signals. In this paper, we theoretically analyze the effects of the proposed methods on the steady-state performance of the system. Also, simulation results are presented and compared to theoretical results. It is concluded that the introduced filter spectral images do not significantly contribute to the error. Rather, the performance is mainly limited by in-band aliasing.