InterSpeech 2021

A Deep Learning Approach to Multi-Channel and Multi-Microphone Acoustic Echo Cancellation
(3 minutes introduction)

Hao Zhang (Ohio State University, USA), DeLiang Wang (Ohio State University, USA)
Building on deep learning based acoustic echo cancellation (AEC) in the single-loudspeaker (single-channel) and single-microphone setup, this paper investigates multi-channel (multi-loudspeaker) AEC (MCAEC) and multi-microphone AEC (MMAEC). A convolutional recurrent network (CRN) is trained to predict the near-end speech from microphone signals with far-end signals used as additional information. We find that the deep learning based MCAEC approach avoids the non-uniqueness problem in traditional MCAEC algorithms. For the AEC setup with multiple microphones, rather than employing AEC for each microphone, we propose to train a single network to achieve echo removal for all microphones. Combining deep learning based AEC with supervised beamforming further improves the system performance. Experimental results show the effectiveness of deep learning approach to MCAEC and MMAEC. Furthermore, deep learning based methods are capable of removing echo and noise simultaneously and work well in the presence of nonlinear distortions.