The USTC-NERCSLIP Systems for The ICMC-ASR Challenge

TL;DR

The paper tackles robust ASR in the ICMC-ASR challenge, where multi-speaker overlap and Mandarin accent dynamics are challenging. It presents an integrated pipeline combining front-end guided source separation with MVDR beamforming, a large-scale semi-supervised data expansion via pseudo-label generation on fusion encoders, a multi-speaker diarization system that fuses SSLR-based x-vectors with MC-TS-VAD, and an Accent-ASR framework that jointly models pronunciation and linguistic information. Key contributions include effective interference suppression in the front-end, scalable PLG-based data augmentation, and accent-aware ASR that improves recognition under varying accents, achieving state-of-the-art results on Track 1 and Track 2. The results demonstrate the practical impact of combining advanced front-end processing, semi-supervised learning, and accent-aware modeling for real-world multi-speaker Mandarin ASR.

Abstract

This report describes the submitted system to the In-Car Multi-Channel Automatic Speech Recognition (ICMC-ASR) challenge, which considers the ASR task with multi-speaker overlapping and Mandarin accent dynamics in the ICMC case. We implement the front-end speaker diarization using the self-supervised learning representation based multi-speaker embedding and beamforming using the speaker position, respectively. For ASR, we employ an iterative pseudo-label generation method based on fusion model to obtain text labels of unsupervised data. To mitigate the impact of accent, an Accent-ASR framework is proposed, which captures pronunciation-related accent features at a fine-grained level and linguistic information at a coarse-grained level. On the ICMC-ASR eval set, the proposed system achieves a CER of 13.16% on track 1 and a cpCER of 21.48% on track 2, which significantly outperforms the official baseline system and obtains the first rank on both tracks.

Figures (1)

• Figure 1: The proposed Accent-ASR framework.