Presentation | 2015-03-02 Modulation Spectrum-Constrained Trajectory Training Algorithm for Statistical Parametric Speech Synthesis Shinnosuke TAKAMICHI, Tomoki TODA, Alan W. BLACK, Satoshi NAKAMURA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper presents a novel training algorithm for statistical parametric speech synthesis. To improve the synthetic speech quality, we have demonstrated the quality improvements by Modulation Spectrum (MS) compensation in synthesis stage. However, such compensation is unsuitable for the systems that need computationally efficient synthesis. The proposed training algorithm is capable of optimizing statistical models considering both a conventional trajectory constraint and a novel MS constraint, and making it possible to (1) use a consistent optimization criterion between training and synthesis processes, (2) use context-dependent modeling for the MS, and (3) provide a closed form solution for the synthesis and compensation processes. The experimental results demonstrate that the proposed algorithm yields significant improvements in speech quality. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | statistical parametric speech synthesis / HMM-based text-to-speech synthesis / MM-based voice conversion / over-smoothing / global variance / modulation spectrum / trajectory model |
Paper # | EA2014-77,SIP2014-118,SP2014-140 |
Date of Issue |
Conference Information | |
Committee | EA |
---|---|
Conference Date | 2015/2/23(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 | Engineering Acoustics (EA) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Modulation Spectrum-Constrained Trajectory Training Algorithm for Statistical Parametric Speech Synthesis |
Sub Title (in English) | |
Keyword(1) | statistical parametric speech synthesis |
Keyword(2) | HMM-based text-to-speech synthesis |
Keyword(3) | MM-based voice conversion |
Keyword(4) | over-smoothing |
Keyword(5) | global variance |
Keyword(6) | modulation spectrum |
Keyword(7) | trajectory model |
1st Author's Name | Shinnosuke TAKAMICHI |
1st Author's Affiliation | Graduate School of Information Science, Nara Institute of Science and Technology (NAIST):Language Technologies Institute, Carnegie Mellon University (CMU)() |
2nd Author's Name | Tomoki TODA |
2nd Author's Affiliation | Language Technologies Institute, Carnegie Mellon University (CMU) |
3rd Author's Name | Alan W. BLACK |
3rd Author's Affiliation | Graduate School of Information Science, Nara Institute of Science and Technology (NAIST) |
4th Author's Name | Satoshi NAKAMURA |
4th Author's Affiliation | Graduate School of Information Science, Nara Institute of Science and Technology (NAIST) |
Date | 2015-03-02 |
Paper # | EA2014-77,SIP2014-118,SP2014-140 |
Volume (vol) | vol.114 |
Number (no) | 473 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |