Title :
Nonlinear scale decomposition based features for visual speech recognition
Author :
Matthews, Iain ; Bangham, J. Andrew ; Harvey, Richard ; Cox, Stephen
Author_Institution :
Sch. of Inf. Syst., Univ. of East Anglia, Norwich, UK
Abstract :
A mathematical morphology based filter structure called a sieve is used to process mouth image sequences of a talker´s mouth and form visual speech features. The effects of varying the type of filter, the post-processing and hidden Markov model (HMM) parameters on recognition accuracy are investigated using two audio-visual speech databases.
Keywords :
audio-visual systems; hidden Markov models; image filtering; image sequences; mathematical morphology; nonlinear filters; speech recognition; HMM parameters; audio-visual speech databases; hidden Markov model; mathematical morphology; mouth image sequence processing; nonlinear scale decomposition; sieve filter structure; visual speech feature recognition; Databases; Hidden Markov models; Histograms; Mouth; Speech; Speech recognition; Visualization;
Conference_Titel :
Signal Processing Conference (EUSIPCO 1998), 9th European
Conference_Location :
Rhodes
Print_ISBN :
978-960-7620-06-4
