Presentation 2012-02-10
Large Vocabrary Sign Language Recognition Based on Cheremes
Arata SATO, Koichi SHINODA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose a large vocabulary sign language recognition method. While most conventional methods used word units for sign language modeling, it is difficult to apply such methods to large-vocabulary recognition. Inspired by the use of phoneme units in speech recognition, we introduce "chereme" unit models in large-vocabulary sign language recognition. The problems are how to define cheremes and developing a recognition algorithm using cheremes. We first represent signs using the 6 elements: "hand movement-hold", "hand shape" and "wrist direction" of both hands. Then, we define chereme unit HMMs in each element. We use the weighted sum of output probabilities of the HMMs of these elements for recognition. We evaluated our method by using isolated sign language recognition. We achieved 78.2% word correct on the closed set of 258 words and 23.6% word correct on the open set with 469 words. We also achieved 18.8% word correct on speaker-independent recognition.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) large-vocabrary sign language recognition / chereme / hidden Markov model
Paper # PRMU2011-222,SP2011-137
Date of Issue

Conference Information
Committee SP
Conference Date 2012/2/2(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Speech (SP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Large Vocabrary Sign Language Recognition Based on Cheremes
Sub Title (in English)
Keyword(1) large-vocabrary sign language recognition
Keyword(2) chereme
Keyword(3) hidden Markov model
1st Author's Name Arata SATO
1st Author's Affiliation Department of Computer Science, Tokyo Institute of Technology()
2nd Author's Name Koichi SHINODA
2nd Author's Affiliation Department of Computer Science, Tokyo Institute of Technology
Date 2012-02-10
Paper # PRMU2011-222,SP2011-137
Volume (vol) vol.111
Number (no) 431
Page pp.pp.-
#Pages 6
Date of Issue