Large-Scale Visual Speech Recognition

TL;DR

This paper introduces a scalable visual speech recognition system trained on the largest dataset to date, significantly outperforming previous lipreading methods in accuracy.

Contribution

The work presents the creation of the largest visual speech dataset and an integrated lipreading system that achieves state-of-the-art accuracy in open-vocabulary speech recognition.

Findings

Achieved 40.9% WER on the test set.

Outperformed professional lipreaders and previous models.

Constructed the largest visual speech dataset with 3,886 hours of video.

Abstract

This work presents a scalable solution to open-vocabulary visual speech recognition. To achieve this, we constructed the largest existing visual speech recognition dataset, consisting of pairs of text and video clips of faces speaking (3,886 hours of video). In tandem, we designed and trained an integrated lipreading system, consisting of a video processing pipeline that maps raw video to stable videos of lips and sequences of phonemes, a scalable deep neural network that maps the lip videos to sequences of phoneme distributions, and a production-level speech decoder that outputs sequences of words. The proposed system achieves a word error rate (WER) of 40.9% as measured on a held-out set. In comparison, professional lipreaders achieve either 86.4% or 92.9% WER on the same dataset when having access to additional types of contextual information. Our approach significantly improves on other lipreading approaches, including variants of LipNet and of Watch, Attend, and Spell (WAS), which are only capable of 89.8% and 76.8% WER respectively.