Presentation 2001/8/24
Analysis of Japanese vowel series by using sandglass type neural network
Eiji KUNIMITSU, Toshie NAMIKI, Hiroki YOSHIMURA, Tadaaki SHIMIZU, Naoki ISU, Kazuhiro SUGATA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The feature of the glide must be investigated for speech synthesis. In order to synthesize high quality speech, the synthesis system is required to reproduce the glide precisely. We proposed a new scheme to derive the characteristics of vowels from LSP parameters as well as compress information by using five-layer nonlinear sandglass type neural network. In our method, 1)SNN learns LSP parameters derived from the five vowels of isolated utterance. 2)The SNN transforms LSP parameters of vowel series to the nonlinear principal component. 3)The feature of the glide is analyzed as locus on the principal component plane.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) snad-glass type neural network / vowel / LSP analysis / formant / principal component analysis
Paper # SP2001-51
Date of Issue

Conference Information
Committee SP
Conference Date 2001/8/24(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) Analysis of Japanese vowel series by using sandglass type neural network
Sub Title (in English)
Keyword(1) snad-glass type neural network
Keyword(2) vowel
Keyword(3) LSP analysis
Keyword(4) formant
Keyword(5) principal component analysis
1st Author's Name Eiji KUNIMITSU
1st Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering()
2nd Author's Name Toshie NAMIKI
2nd Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering
3rd Author's Name Hiroki YOSHIMURA
3rd Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering
4th Author's Name Tadaaki SHIMIZU
4th Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering
5th Author's Name Naoki ISU
5th Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering
6th Author's Name Kazuhiro SUGATA
6th Author's Affiliation Tottori University, Faculty of engineering, Department of information and knowledge engineering
Date 2001/8/24
Paper # SP2001-51
Volume (vol) vol.101
Number (no) 271
Page pp.pp.-
#Pages 8
Date of Issue