Presentation 2002/8/23
Robustness of Automatic Labeling with HMM Phoneme Models
Tomoko YONEZAWA, Hideyuki MIZUNO, Masanobu ABE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) To realize speech synthesis of different speakers and different speech styles, we aim to build a full-automatic phoneme labeling system which sorts out the database of the speech synthesis units including various speakers and speech styles. In Preparation, we experimented automatic labeling with HMM phoneme models from Mel Filter Bank and Mel Cepstrum. First we observed the results with high accuracy of label boundaries from Mel Filter Bank's ten channels. Then we analyzed the results by categorizing them into several chains of phonemes. Second we observed that there are different accuracy among different phoneme chains. Additionally, we observed robust accuracy in some chains depending upon the speakers and speech styles.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) HMM / Phoneme Models / Automatic Labeling / Mel Filter Bank / Mel Cepstrum
Paper # SP2002-74
Date of Issue

Conference Information
Committee SP
Conference Date 2002/8/23(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) Robustness of Automatic Labeling with HMM Phoneme Models
Sub Title (in English)
Keyword(1) HMM
Keyword(2) Phoneme Models
Keyword(3) Automatic Labeling
Keyword(4) Mel Filter Bank
Keyword(5) Mel Cepstrum
1st Author's Name Tomoko YONEZAWA
1st Author's Affiliation NTT Cyber Space Laboratories, NTT Corporation()
2nd Author's Name Hideyuki MIZUNO
2nd Author's Affiliation NTT Cyber Space Laboratories, NTT Corporation
3rd Author's Name Masanobu ABE
3rd Author's Affiliation NTT Cyber Space Laboratories, NTT Corporation
Date 2002/8/23
Paper # SP2002-74
Volume (vol) vol.102
Number (no) 292
Page pp.pp.-
#Pages 6
Date of Issue