Presentation 2002/11/15
Automatic Lipreading of Station Names Using Optical Flow and HMM
Takashi OTSUKI, Teruhiko OHTOMO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this report, we present an automatic lipreading system which uses optical flow as feature vector, and word HMMs for recognition. The feature vectors in this system are smoothed optical flow extracted by the image processing hardware "Color Tracking Vision". The feature vectors are converted to symbols by vector quantization, then the symbol series are fed to the HMM word recognition unit. The word recognition experiment of 50 railroad station names shows that the optimal structure of feature vector, and the optimal number of states in word HMM. Using position-perturbed data for learning, very high recognition accuracy and very high tolerance against misalignment of mouse position are accomplished.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) automatic lipreading / image recognition / optical flow / color Tracking Vision / VQ / HMM
Paper # HIP2002-31
Date of Issue

Conference Information
Committee HIP
Conference Date 2002/11/15(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Vice Chair

Paper Information
Registration To Human Information Processing (HIP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Automatic Lipreading of Station Names Using Optical Flow and HMM
Sub Title (in English)
Keyword(1) automatic lipreading
Keyword(2) image recognition
Keyword(3) optical flow
Keyword(4) color Tracking Vision
Keyword(5) VQ
Keyword(6) HMM
1st Author's Name Takashi OTSUKI
1st Author's Affiliation ()
2nd Author's Name Teruhiko OHTOMO
2nd Author's Affiliation
Date 2002/11/15
Paper # HIP2002-31
Volume (vol) vol.102
Number (no) 473
Page pp.pp.-
#Pages 6
Date of Issue