Leveraging Real Conversational Data for Multi-Channel Continuous Speech Separation

TL;DR

This paper introduces a three-stage training scheme for multi-channel continuous speech separation that effectively utilizes both supervised and large-scale unsupervised real conversational data, improving meeting transcription accuracy.

Contribution

A novel semi-supervised training approach combining simulated, transcribed, and real data with teacher-student learning for CSS models.

Findings

Steady performance improvements at each training stage.

Effective leveraging of real conversational data.

Enhanced multi-channel CSS for meeting transcription.

Abstract

Existing multi-channel continuous speech separation (CSS) models are heavily dependent on supervised data - either simulated data which causes data mismatch between the training and real-data testing, or the real transcribed overlapping data, which is difficult to be acquired, hindering further improvements in the conversational/meeting transcription tasks. In this paper, we propose a three-stage training scheme for the CSS model that can leverage both supervised data and extra large-scale unsupervised real-world conversational data. The scheme consists of two conventional training approaches -- pre-training using simulated data and ASR-loss-based training using transcribed data -- and a novel continuous semi-supervised training between the two, in which the CSS model is further trained by using real data based on the teacher-student learning framework. We apply this scheme to an array-geometry-agnostic CSS model, which can use the multi-channel data collected from any microphone array. Large-scale meeting transcription experiments are carried out on both Microsoft internal meeting data and the AMI meeting corpus. The steady improvement by each training stage has been observed, showing the effect of the proposed method that enables leveraging real conversational data for CSS model training.