Output details
11 - Computer Science and Informatics
Queen's University Belfast
Comparison of Image Transform-Based Features for Visual Speech Recognition in Clean and Corrupted Videos
<23> This paper is the first comprehensive investigation of image transform features for visual speech recognition. It shows that novel combinations of features from different transforms can provide robust performance in real-world conditions. The success of this project was a significant factor in gaining SBRI funding from TSB (the top ranked proposal out of 60 in UK) to develop a prototype system which forms the basis for a spin out company (Liopa).The associated business plan won the “digital media and software” category in the NISP 25K competition. We have also been invited to apply for phase 2 SBRI funding.