Enhancing CTC-Based Visual Speech Recognition

TL;DR

LiteVSR2 is an improved CTC-based visual speech recognition model that achieves higher accuracy through stabilized preprocessing and feature normalization, without extra data or computational cost.

Contribution

The paper introduces LiteVSR2, a novel, resource-efficient VSR model with enhanced performance achieved via stabilized preprocessing and feature normalization techniques.

Findings

Achieves state-of-the-art results on LRS2 and LRS3 benchmarks.

Maintains efficiency while significantly improving accuracy.

Scales effectively across different model sizes and data volumes.

Abstract

This paper presents LiteVSR2, an enhanced version of our previously introduced efficient approach to Visual Speech Recognition (VSR). Building upon our knowledge distillation framework from a pre-trained Automatic Speech Recognition (ASR) model, we introduce two key improvements: a stabilized video preprocessing technique and feature normalization in the distillation process. These improvements yield substantial performance gains on the LRS2 and LRS3 benchmarks, positioning LiteVSR2 as the current best CTC-based VSR model without increasing the volume of training data or computational resources utilized. Furthermore, we explore the scalability of our approach by examining performance metrics across varying model complexities and training data volumes. LiteVSR2 maintains the efficiency of its predecessor while significantly enhancing accuracy, thereby demonstrating the potential for resource-efficient advancements in VSR technology.