Posterior Residual Echo Canceling and its Complexity Reduction in the Wave Domain

Abstract:
Better echo cancellation is desirable for full-duplex immersive teleconferencing system using massive multi-channel reproduction techniques such as wave field synthesis (WFS). Instead of point-to-point channel identification, wave-domain adaptive filtering was proposed to reduce the complexity of the loudspeaker enclosure microphone (LEM) system drastically. This LEM system model in the wave domain was generalized to improve the echo cancellation when multiple sound sources are reproduced at the same time. For the same purpose, we propose here a posterior wave-domain residual echo canceling method. We further show that we can reduce the computational complexity of this residual echo canceling by exploiting the property of the sound field after two dimensional (2D) Fourier transform. Experimental evaluations support the applicability of the proposed methods. Index Terms - wave domain, adaptive filtering, residual echo, 2D spectrum