Transformer-Based Video Front-Ends for Audio-Visual Speech Recognition for Single and Multi-Person Video

TL;DR

This paper introduces a transformer-based video front-end for audio-visual speech recognition, replacing traditional 3D CNNs, leading to improved accuracy on large-scale YouTube and LRS3-TED datasets, including multi-person scenarios.

Contribution

The work proposes using video transformer networks instead of 3D CNNs for visual feature extraction in AV-ASR, achieving state-of-the-art results.

Findings

Achieved 17.0% WER on LRS3-TED with the proposed model.

Improved recognition accuracy by 10-15% over convolutional baselines.

Reduced error rates in multi-person AV-ASR scenarios.

Abstract

Audio-visual automatic speech recognition (AV-ASR) extends speech recognition by introducing the video modality as an additional source of information. In this work, the information contained in the motion of the speaker's mouth is used to augment the audio features. The video modality is traditionally processed with a 3D convolutional neural network (e.g. 3D version of VGG). Recently, image transformer networks arXiv:2010.11929 demonstrated the ability to extract rich visual features for image classification tasks. Here, we propose to replace the 3D convolution with a video transformer to extract visual features. We train our baselines and the proposed model on a large scale corpus of YouTube videos. The performance of our approach is evaluated on a labeled subset of YouTube videos as well as on the LRS3-TED public corpus. Our best video-only model obtains 31.4% WER on YTDEV18 and 17.0% on LRS3-TED, a 10% and 15% relative improvements over our convolutional baseline. We achieve the state of the art performance of the audio-visual recognition on the LRS3-TED after fine-tuning our model (1.6% WER). In addition, in a series of experiments on multi-person AV-ASR, we obtained an average relative reduction of 2% over our convolutional video frontend.