End-to-end training of time domain audio separation and recognition

TL;DR

This paper presents a novel end-to-end model combining time domain speech separation with recognition, achieving significant WER improvements on multi-speaker datasets by joint training of separation and recognition modules.

Contribution

It introduces a joint training framework for time domain separation and recognition, bridging a gap in multi-speaker speech processing research.

Findings

Achieved 11.0% WER on WSJ0-2mix dataset.

Demonstrated substantial improvements over previous cascade and frequency domain models.

Abstract

The rising interest in single-channel multi-speaker speech separation sparked development of End-to-End (E2E) approaches to multi-speaker speech recognition. However, up until now, state-of-the-art neural network-based time domain source separation has not yet been combined with E2E speech recognition. We here demonstrate how to combine a separation module based on a Convolutional Time domain Audio Separation Network (Conv-TasNet) with an E2E speech recognizer and how to train such a model jointly by distributing it over multiple GPUs or by approximating truncated back-propagation for the convolutional front-end. To put this work into perspective and illustrate the complexity of the design space, we provide a compact overview of single-channel multi-speaker recognition systems. Our experiments show a word error rate of 11.0% on WSJ0-2mix and indicate that our joint time domain model can yield substantial improvements over cascade DNN-HMM and monolithic E2E frequency domain systems proposed so far.