VATLM: Visual-Audio-Text Pre-Training with Unified Masked Prediction for Speech Representation Learning
This work addresses the challenge of multimodal speech interaction for applications like speech recognition, though it appears incremental as it builds on existing pre-training methods.
The paper tackles the problem of integrating multimodal information (vision, audio, text) for speech representation learning by proposing VATLM, a unified framework that uses masked prediction to align modalities, and results show it outperforms state-of-the-art models like AV-HuBERT on tasks such as audio-visual speech recognition.
Although speech is a simple and effective way for humans to communicate with the outside world, a more realistic speech interaction contains multimodal information, e.g., vision, text. How to design a unified framework to integrate different modal information and leverage different resources (e.g., visual-audio pairs, audio-text pairs, unlabeled speech, and unlabeled text) to facilitate speech representation learning was not well explored. In this paper, we propose a unified cross-modal representation learning framework VATLM (Visual-Audio-Text Language Model). The proposed VATLM employs a unified backbone network to model the modality-independent information and utilizes three simple modality-dependent modules to preprocess visual, speech, and text inputs. In order to integrate these three modalities into one shared semantic space, VATLM is optimized with a masked prediction task of unified tokens, given by our proposed unified tokenizer. We evaluate the pre-trained VATLM on audio-visual related downstream tasks, including audio-visual speech recognition (AVSR), visual speech recognition (VSR) tasks. Results show that the proposed VATLM outperforms previous the state-of-the-art models, such as audio-visual pre-trained AV-HuBERT model, and analysis also demonstrates that VATLM is capable of aligning different modalities into the same space. To facilitate future research, we release the code and pre-trained models at https://aka.ms/vatlm.