Presentation 1997/12/12
A Study on Word Spotting based on Likelihood Normalization Using Phoneme HMMs
Masaharu KATOH, Takaaki HORI, Akinori ITO, Masaki KOHDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In recent speech recognition, hidden Markov model (HMM) has been useful. We consider likelihood score of HMMs from a point of theory of probability. In continuous speech recognition, each hypothesis will have different length and position of speech segment. It affects the system performance by comparing the HMMs' scores directly. In this paper, we describe normalization of likelihood based on Bayes' theorem. To normalize likelihood, we use connected phoneme HMMs that allow Japanese syllable rule. In this method, we need no additional calculation to get scores, and we need no models except phoneme HMMs to the system. We apply it to the word-spotting, and obtain significant improvement of system performance.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) normalization of likelihood / word spotting / score function / connected phoneme HMMs
Paper # NLC97-44
Date of Issue

Conference Information
Committee NLC
Conference Date 1997/12/12(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 Natural Language Understanding and Models of Communication (NLC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Study on Word Spotting based on Likelihood Normalization Using Phoneme HMMs
Sub Title (in English)
Keyword(1) normalization of likelihood
Keyword(2) word spotting
Keyword(3) score function
Keyword(4) connected phoneme HMMs
1st Author's Name Masaharu KATOH
1st Author's Affiliation Yamagata University()
2nd Author's Name Takaaki HORI
2nd Author's Affiliation Yamagata University
3rd Author's Name Akinori ITO
3rd Author's Affiliation Yamagata University
4th Author's Name Masaki KOHDA
4th Author's Affiliation Yamagata University
Date 1997/12/12
Paper # NLC97-44
Volume (vol) vol.97
Number (no) 440
Page pp.pp.-
#Pages 6
Date of Issue