Multi-channel Acoustic Echo Cancellation Based on Residual Echo Enhancement with Effective Channel Decorrelation via Resampling

The residual echo enhancement (REE) procedure proposed in [1] is able to improve the acoustic echo cancellation (AEC) performance in a very noisy acoustic mixing environment by utilizing the natural learning ability of the least-mean square (LMS) algorithm without precise estimation of the signal statistics. We demonstrate in this paper that the technique can also be applied effectively in a multichannel AEC (MCAEC) setting to indirectly assist in the recovery of lost AEC performance due to the non-uniqueness problem. In addition, we incorporate other techniques to further boost the REE-based MCAEC performance. One of the techniques is a new channel decorrelation procedure based on resampling that directly alleviates the non-uniqueness problem while introducing minimal distortion to signal quality and statistics.