Exploring End-to-End Multi-channel ASR with Bias Information for Meeting Transcription

TL;DR

This paper investigates joint multi-channel speech enhancement and recognition for meeting transcription, leveraging large-scale single-channel data and bias information to improve accuracy in practical scenarios.

Contribution

It introduces a novel joint modeling framework with effective training strategies and a location bias integration method, enhancing meeting transcription performance.

Findings

Significant WER reduction achieved on meeting recordings.

Effective use of simulated and real multi-channel data for training.

Proposed deep concatenation method improves location bias integration.

Abstract

Joint optimization of multi-channel front-end and automatic speech recognition (ASR) has attracted much interest. While promising results have been reported for various tasks, past studies on its meeting transcription application were limited to small scale experiments. It is still unclear whether such a joint framework can be beneficial for a more practical setup where a massive amount of single channel training data can be leveraged for building a strong ASR back-end. In this work, we present our investigation on the joint modeling of a mask-based beamformer and Attention-Encoder-Decoder-based ASR in the setting where we have 75k hours of single-channel data and a relatively small amount of real multi-channel data for model training. We explore effective training procedures, including a comparison of simulated and real multi-channel training data. To guide the recognition towards a target speaker and deal with overlapped speech, we also explore various combinations of bias information, such as direction of arrivals and speaker profiles. We propose an effective location bias integration method called deep concatenation for the beamformer network. In our evaluation on various meeting recordings, we show that the proposed framework achieves a substantial word error rate reduction.