Enhancing CTC-Based Visual Speech Recognition
This work addresses resource-efficient advancements in VSR technology, which is important for applications like assistive devices or noisy environments, though it is incremental as it builds directly on a previous method.
The paper tackles improving visual speech recognition (VSR) by enhancing an existing CTC-based model with stabilized video preprocessing and feature normalization, achieving state-of-the-art results on LRS2 and LRS3 benchmarks without additional data or computational resources.
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.