Where Visual Speech Meets Language: VSP-LLM Framework for Efficient and Context-Aware Visual Speech Processing
This work addresses the challenge of homophenes in visual speech recognition and translation for applications like assistive technology, though it appears incremental by combining existing methods like self-supervised models and LoRA.
The paper tackles the problem of ambiguous lip movements in visual speech processing by proposing the VSP-LLM framework, which integrates LLMs for context-aware multi-task recognition and translation, achieving more effective translation on the MuAViC benchmark with only 30 hours of labeled data compared to a model trained on 433 hours.
In visual speech processing, context modeling capability is one of the most important requirements due to the ambiguous nature of lip movements. For example, homophenes, words that share identical lip movements but produce different sounds, can be distinguished by considering the context. In this paper, we propose a novel framework, namely Visual Speech Processing incorporated with LLMs (VSP-LLM), to maximize the context modeling ability by bringing the overwhelming power of LLMs. Specifically, VSP-LLM is designed to perform multi-tasks of visual speech recognition and translation, where the given instructions control the type of task. The input video is mapped to the input latent space of an LLM by employing a self-supervised visual speech model. Focused on the fact that there is redundant information in input frames, we propose a novel deduplication method that reduces the embedded visual features by employing visual speech units. Through the proposed deduplication and Low Rank Adaptation (LoRA), VSP-LLM can be trained in a computationally efficient manner. In the translation dataset, the MuAViC benchmark, we demonstrate that VSP-LLM trained on just 30 hours of labeled data can more effectively translate lip movements compared to the recent model trained with 433 hours of data.