End-to-End Far-Field Speech Recognition with Unified Dereverberation and Beamforming

TL;DR

This paper presents an integrated end-to-end speech recognition system that combines dereverberation and beamforming, improving performance in reverberant environments through novel frontend architectures optimized with speech recognition criteria.

Contribution

It introduces two new frontend architectures for dereverberation within end-to-end speech recognition, including a multi-source WPE and an extended WPD convolutional beamformer, with improved stability and performance.

Findings

Outperforms conventional methods in reverberant scenarios

Effective integration of dereverberation into end-to-end systems

Stable back-propagation with matrix inverse in WPD extension

Abstract

Despite successful applications of end-to-end approaches in multi-channel speech recognition, the performance still degrades severely when the speech is corrupted by reverberation. In this paper, we integrate the dereverberation module into the end-to-end multi-channel speech recognition system and explore two different frontend architectures. First, a multi-source mask-based weighted prediction error (WPE) module is incorporated in the frontend for dereverberation. Second, another novel frontend architecture is proposed, which extends the weighted power minimization distortionless response (WPD) convolutional beamformer to perform simultaneous separation and dereverberation. We derive a new formulation from the original WPD, which can handle multi-source input, and replace eigenvalue decomposition with the matrix inverse operation to make the back-propagation algorithm more stable. The above two architectures are optimized in a fully end-to-end manner, only using the speech recognition criterion. Experiments on both spatialized wsj1-2mix corpus and REVERB show that our proposed model outperformed the conventional methods in reverberant scenarios.