ViCocktail: Automated Multi-Modal Data Collection for Vietnamese Audio-Visual Speech Recognition

TL;DR

This paper introduces ViCocktail, an automated multi-modal data collection method for Vietnamese AVSR, improving dataset creation efficiency and enabling robust speech recognition in noisy environments for under-resourced languages.

Contribution

It presents a practical automated data collection approach for AVSR datasets, specifically tailored for Vietnamese, enhancing dataset accessibility and model robustness.

Findings

Achieved competitive AVSR performance in clean conditions.

Significantly outperformed traditional ASR in noisy environments.

Demonstrated broad applicability to under-resourced languages.

Abstract

Audio-Visual Speech Recognition (AVSR) has gained significant attention recently due to its robustness against noise, which often challenges conventional speech recognition systems that rely solely on audio features. Despite this advantage, AVSR models remain limited by the scarcity of extensive datasets, especially for most languages beyond English. Automated data collection offers a promising solution. This work presents a practical approach to generate AVSR datasets from raw video, refining existing techniques for improved efficiency and accessibility. We demonstrate its broad applicability by developing a baseline AVSR model for Vietnamese. Experiments show the automatically collected dataset enables a strong baseline, achieving competitive performance with robust ASR in clean conditions and significantly outperforming them in noisy environments like cocktail parties. This efficient method provides a pathway to expand AVSR to more languages, particularly under-resourced ones.