Which phoneme-to-viseme maps best improve visual-only computer lip-reading?
This work addresses a critical assumption in visual speech recognition systems, providing incremental improvements for researchers and developers in automated lip-reading.
The paper tackled the problem of identifying effective phoneme-to-viseme mappings for visual-only lip-reading by testing 120 existing maps and developing new ones based on phoneme confusions, resulting in improved mappings for individual talkers.
A critical assumption of all current visual speech recognition systems is that there are visual speech units called visemes which can be mapped to units of acoustic speech, the phonemes. Despite there being a number of published maps it is infrequent to see the effectiveness of these tested, particularly on visual-only lip-reading (many works use audio-visual speech). Here we examine 120 mappings and consider if any are stable across talkers. We show a method for devising maps based on phoneme confusions from an automated lip-reading system, and we present new mappings that show improvements for individual talkers.