Interactive decoding of words from visual speech recognition models
This addresses the challenge of improving accuracy in visual speech recognition systems for users, though it appears incremental as it builds on existing phoneme-to-word pipelines.
The paper tackles the problem of ambiguity in visual speech recognition by introducing an interactive decoding method that allows users to direct the decoding process at each word position, showing promise for text input applications through automated oracle-based evaluation.
This work describes an interactive decoding method to improve the performance of visual speech recognition systems using user input to compensate for the inherent ambiguity of the task. Unlike most phoneme-to-word decoding pipelines, which produce phonemes and feed these through a finite state transducer, our method instead expands words in lockstep, facilitating the insertion of interaction points at each word position. Interaction points enable us to solicit input during decoding, allowing users to interactively direct the decoding process. We simulate the behavior of user input using an oracle to give an automated evaluation, and show promise for the use of this method for text input.