Presentation 2012-01-19
Binarizing threshold control for Distance Matrix Images in telephone conversations retrieval
Hiroyuki Nishi, Yuuki Yokobayashi, Tran Li Haiyen, Yoshimasa Kimura, Toshio Kakinoki,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Telephone conversation retrieval systems are required for mail order, help desk and similar telecommunications services. The researches of spoken term detection drove using strong language models as well as conventional acoustic model. However it is difficult to construct the language model of natural telephone conversations. On the other hand, key entry method by speech is also required. In order to solve these problems, word spotting can be used. However, conventional word spotting, for exsample, continuous dynamic programming (DP) can not be appled because its accuracy is not enough for such applications. This report proposes a new method in order to realize telephone conversation log retrieval. The method is structured by distance matrix images (DMI) and the line detection in the images. Important parameters are threshold values for binalizing of DMI. The merit of this method is not to need a dictionary and the detailed language model. This report shows the novel threshold level control method using the stchastic characteristics of DMI.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) speech recognition / word spotting / life log / telephony / conversation log
Paper # ICM2011-32,LOIS2011-57
Date of Issue

Conference Information
Committee LOIS
Conference Date 2012/1/12(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 Life Intelligence and Office Information Systems (LOIS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Binarizing threshold control for Distance Matrix Images in telephone conversations retrieval
Sub Title (in English)
Keyword(1) speech recognition
Keyword(2) word spotting
Keyword(3) life log
Keyword(4) telephony
Keyword(5) conversation log
1st Author's Name Hiroyuki Nishi
1st Author's Affiliation Faculty of Information Sciences, Sojo University()
2nd Author's Name Yuuki Yokobayashi
2nd Author's Affiliation Faculty of Information Sciences, Sojo University
3rd Author's Name Tran Li Haiyen
3rd Author's Affiliation Faculty of Information Sciences, Sojo University
4th Author's Name Yoshimasa Kimura
4th Author's Affiliation Faculty of Information Sciences, Sojo University
5th Author's Name Toshio Kakinoki
5th Author's Affiliation Faculty of Information Sciences, Sojo University
Date 2012-01-19
Paper # ICM2011-32,LOIS2011-57
Volume (vol) vol.111
Number (no) 383
Page pp.pp.-
#Pages 6
Date of Issue