Symposium 25
Symposium 25

Edward Kmett

Department of Mathematics

Viseme Extraction Versus Speech Recognition

A mathematical overview of statistical techniques used for extracting “lip-sync” information from an audio stream will be investigated. These statistical techniques differ from conventional approaches derived from speech recognition technology in that they can be used in real-time and do not require the entire utterance to be present before speech recognition processing can begin.

Poster - 10:30 a.m.~12:30 p.m. - Guild Hall.

Eastern Michigan University