Two-step Band-split Neural Network Approach for Full-band Residual Echo Suppression

TL;DR

This paper introduces a two-step neural network method for full-band residual echo suppression, splitting signals into wide and high bands for targeted processing, achieving high MOS scores and ranking second in a challenge.

Contribution

The novel two-step band-split neural network approach effectively handles full-band residual echo suppression with improved accuracy and reduced complexity.

Findings

Achieved MOS of 4.344 on ICASSP 2023 AEC Challenge

Ranked 2nd (tied) in the non-personalized track

Effective separation of wide-band and high-band signals

Abstract

This paper describes a Two-step Band-split Neural Network (TBNN) approach for full-band acoustic echo cancellation. Specifically, after linear filtering, we split the full-band signal into wide-band (16KHz) and high-band (16-48KHz) for residual echo removal with lower modeling difficulty. The wide-band signal is processed by an updated gated convolutional recurrent network (GCRN) with U$^2$ encoder while the high-band signal is processed by a high-band post-filter net with lower complexity. Our approach submitted to ICASSP 2023 AEC Challenge has achieved an overall mean opinion score (MOS) of 4.344 and a word accuracy (WAcc) ratio of 0.795, leading to the 2$^{nd}$ (tied) in the ranking of the non-personalized track.