NeuralEcho: A Self-Attentive Recurrent Neural Network For Unified Acoustic Echo Suppression And Speech Enhancement

TL;DR

NeuralEcho introduces a self-attentive recurrent neural network that jointly performs acoustic echo suppression and speech enhancement, outperforming existing methods in quality, recognition accuracy, and complexity, while supporting speaker embedding and automatic gain control.

Contribution

It presents a novel all-deep-learning framework that implicitly estimates echo/noise and speech statistics for unified suppression and enhancement, advancing the state-of-the-art.

Findings

Outperforms F-T-LSTM in speech quality metrics

Improves speech recognition accuracy

Supports speaker embedding and automatic gain control

Abstract

Acoustic echo cancellation (AEC) plays an important role in the full-duplex speech communication as well as the front-end speech enhancement for recognition in the conditions when the loudspeaker plays back. In this paper, we present an all-deep-learning framework that implicitly estimates the second order statistics of echo/noise and target speech, and jointly solves echo and noise suppression through an attention based recurrent neural network. The proposed model outperforms the state-of-the-art joint echo cancellation and speech enhancement method F-T-LSTM in terms of objective speech quality metrics, speech recognition accuracy and model complexity. We show that this model can work with speaker embedding for better target speech enhancement and furthermore develop a branch for automatic gain control (AGC) task to form an all-in-one front-end speech enhancement system.