Presentation | 2010-01-21 Performance evaluation of Voice Conversion Based on F0 Quantization and Non-parallel Training Yuhei OTA, Takashi NOSE, Takao KOBAYASHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper describes the performance evaluation results of a context-dependent HMM-based voice conversion technique to show its effectiveness by comparing with a GMM-based one. In the HMM-based conversion, first we extract the phonetic and prosodic information from input speech of a source speaker. Then, converted synthetic speech is generated from the pre-trained acoustic model of a target speaker. To appropriately model the pitch information, we use a roughly quantized FO symbol sequence as the prosodic context instead of accent information obtained by manual labeling for training data. By using the phonetically and prosodically context-dependent HMMs, the speaker characteristics appearing in segmental and supra-segmental features can be also converted, which is difficult in conventional GMM-based techniques. Objective and subjective experimental results show that the naturalness and speaker individuality of converted speech are significantly improved by using HMM-based voice conversion. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | voice conversion / HMM-based speech synthesis / prosodic information / F0 quantization / GMM |
Paper # | CQ2009-60,PRMU2009-159,SP2009-100,MVE2009-82 |
Date of Issue |
Conference Information | |
Committee | CQ |
---|---|
Conference Date | 2010/1/14(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 | Communication Quality (CQ) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Performance evaluation of Voice Conversion Based on F0 Quantization and Non-parallel Training |
Sub Title (in English) | |
Keyword(1) | voice conversion |
Keyword(2) | HMM-based speech synthesis |
Keyword(3) | prosodic information |
Keyword(4) | F0 quantization |
Keyword(5) | GMM |
1st Author's Name | Yuhei OTA |
1st Author's Affiliation | Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology() |
2nd Author's Name | Takashi NOSE |
2nd Author's Affiliation | Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology |
3rd Author's Name | Takao KOBAYASHI |
3rd Author's Affiliation | Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology |
Date | 2010-01-21 |
Paper # | CQ2009-60,PRMU2009-159,SP2009-100,MVE2009-82 |
Volume (vol) | vol.109 |
Number (no) | 373 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |