MULTI-TASK SUB-BAND NETWORK FOR DEEP RESIDUAL ECHO SUPPRESSION

This paper introduces the SWANT team’s entry to the ICASSP 2023 AEC Challenge. We submit a system that cascades a linear filter with a neural post-filter. Particularly, we adopt sub-band process- ing to handle full-band signal and shape the network with multi-task learning, where dual signal voice activity detection (DSVAD) and echo estimation are adopted as auxiliary tasks. Moreover, we par- ticularly improve the time frequency convolution module (TFCM) to increase the receptive field using small convolution kernels. Fi- nally, our system has ranked 4th in ICASSP 2023 AEC Challenge Non-personalized track.