NTT Multi-Speaker ASR System for the DASR Task of CHiME-8 Challenge

TL;DR

This paper introduces a comprehensive multi-speaker distant speech recognition system for the CHiME-8 challenge, combining diarization, speaker counting, source separation, and ASR models, achieving significant performance improvements.

Contribution

It presents a novel multi-channel speaker counting method and an integrated pipeline for DASR, combining advanced diarization, source separation, and pre-trained ASR models.

Findings

Achieved 21.3% macro tcpWER on dev set, 57% better than baseline.

Developed a new multi-channel speaker counting approach.

Enhanced source separation with guided source separation (GSS).

Abstract

We present a distant automatic speech recognition (DASR) system developed for the CHiME-8 DASR track. It consists of a diarization first pipeline. For diarization, we use end-to-end diarization with vector clustering (EEND-VC) followed by target speaker voice activity detection (TS-VAD) refinement. To deal with various numbers of speakers, we developed a new multi-channel speaker counting approach. We then apply guided source separation (GSS) with several improvements to the baseline system. Finally, we perform ASR using a combination of systems built from strong pre-trained models. Our proposed system achieves a macro tcpWER of 21.3 % on the dev set, which is a 57 % relative improvement over the baseline.