.

Back...

Naming

Speech motor recognition (or silent speech recognition, articulation reading) means recognizing phonemes/words/sentences from artiuclatory movement data (usually includinng tongue and lip data). It does not mean the speech recognition (from acoustics) with the asistance of articulatory modeling (called audio-visual speech recognition or articulatory speech recognition). The challenges of slient speech reading lie in the complexity of human speech production mechanism, for example, many-to-one mapping between articulatory configurations and speech, co-articulation, and high degree of variation of articulatory movements within and acorss talkers.

Following graph may illustrate the current research fields regarding the relations among articulatory configurations, speech, and acoustics. The brown arrow is the direction of our research.