The USTC-NERCSLIP Systems for the CHiME-8 NOTSOFAR-1 Challenge

TL;DR

This paper presents a robust multi-channel speech processing system for the CHiME-8 challenge, combining joint training for diarization and separation with advanced ASR techniques, achieving state-of-the-art error rates.

Contribution

It introduces a data-driven joint training method for diarization and separation, and enhances Whisper ASR with multiple innovations for improved robustness in real-world conditions.

Findings

Achieved a tcpWER of 14.265% on multi-channel data.

Achieved a tcpWER of 22.989% on single-channel data.

Demonstrated effectiveness of combined GSS and JDS methods.

Abstract

This technical report outlines our submission system for the CHiME-8 NOTSOFAR-1 Challenge. The primary difficulty of this challenge is the dataset recorded across various conference rooms, which captures real-world complexities such as high overlap rates, background noises, a variable number of speakers, and natural conversation styles. To address these issues, we optimized the system in several aspects: For front-end speech signal processing, we introduced a data-driven joint training method for diarization and separation (JDS) to enhance audio quality. Additionally, we also integrated traditional guided source separation (GSS) for multi-channel track to provide complementary information for the JDS. For back-end speech recognition, we enhanced Whisper with WavLM, ConvNeXt, and Transformer innovations, applying multi-task training and Noise KLD augmentation, to significantly advance ASR robustness and accuracy. Our system attained a Time-Constrained minimum Permutation Word Error Rate (tcpWER) of 14.265% and 22.989% on the CHiME-8 NOTSOFAR-1 Dev-set-2 multi-channel and single-channel tracks, respectively.