A Hidden Semi-Markov Model-Based Speech Synthesis System

Heiga ZEN; Keiichi TOKUDA; Takashi MASUKO; Takao KOBAYASIH; Tadashi KITAMURA

doi:10.1093/ietisy/e90-d.5.825

A Hidden Semi-Markov Model-Based Speech Synthesis System

Heiga ZEN, Keiichi TOKUDA, Takashi MASUKO, Takao KOBAYASIH, Tadashi KITAMURA

Full Text Views

0

Share
Cite this

Summary :

A statistical speech synthesis system based on the hidden Markov model (HMM) was recently proposed. In this system, spectrum, excitation, and duration of speech are modeled simultaneously by context-dependent HMMs, and speech parameter vector sequences are generated from the HMMs themselves. This system defines a speech synthesis problem in a generative model framework and solves it based on the maximum likelihood (ML) criterion. However, there is an inconsistency: although state duration probability density functions (PDFs) are explicitly used in the synthesis part of the system, they have not been incorporated into its training part. This inconsistency can make the synthesized speech sound less natural. In this paper, we propose a statistical speech synthesis system based on a hidden semi-Markov model (HSMM), which can be viewed as an HMM with explicit state duration PDFs. The use of HSMMs can solve the above inconsistency because we can incorporate the state duration PDFs explicitly into both the synthesis and the training parts of the system. Subjective listening test results show that use of HSMMs improves the reported naturalness of synthesized speech.

Publication: IEICE TRANSACTIONS on Information Vol.E90-D No.5 pp.825-834

Publication Date: 2007/05/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1093/ietisy/e90-d.5.825

Type of Manuscript: PAPER

Category: Speech and Hearing

Cite this

Copy

Heiga ZEN, Keiichi TOKUDA, Takashi MASUKO, Takao KOBAYASIH, Tadashi KITAMURA, "A Hidden Semi-Markov Model-Based Speech Synthesis System" in IEICE TRANSACTIONS on Information, vol. E90-D, no. 5, pp. 825-834, May 2007, doi: 10.1093/ietisy/e90-d.5.825.
Abstract: A statistical speech synthesis system based on the hidden Markov model (HMM) was recently proposed. In this system, spectrum, excitation, and duration of speech are modeled simultaneously by context-dependent HMMs, and speech parameter vector sequences are generated from the HMMs themselves. This system defines a speech synthesis problem in a generative model framework and solves it based on the maximum likelihood (ML) criterion. However, there is an inconsistency: although state duration probability density functions (PDFs) are explicitly used in the synthesis part of the system, they have not been incorporated into its training part. This inconsistency can make the synthesized speech sound less natural. In this paper, we propose a statistical speech synthesis system based on a hidden semi-Markov model (HSMM), which can be viewed as an HMM with explicit state duration PDFs. The use of HSMMs can solve the above inconsistency because we can incorporate the state duration PDFs explicitly into both the synthesis and the training parts of the system. Subjective listening test results show that use of HSMMs improves the reported naturalness of synthesized speech.
URL: https://globals.ieice.org/en_transactions/information/10.1093/ietisy/e90-d.5.825/_p

Copy

@ARTICLE{e90-d_5_825,
author={Heiga ZEN, Keiichi TOKUDA, Takashi MASUKO, Takao KOBAYASIH, Tadashi KITAMURA, },
journal={IEICE TRANSACTIONS on Information},
title={A Hidden Semi-Markov Model-Based Speech Synthesis System},
year={2007},
volume={E90-D},
number={5},
pages={825-834},
abstract={A statistical speech synthesis system based on the hidden Markov model (HMM) was recently proposed. In this system, spectrum, excitation, and duration of speech are modeled simultaneously by context-dependent HMMs, and speech parameter vector sequences are generated from the HMMs themselves. This system defines a speech synthesis problem in a generative model framework and solves it based on the maximum likelihood (ML) criterion. However, there is an inconsistency: although state duration probability density functions (PDFs) are explicitly used in the synthesis part of the system, they have not been incorporated into its training part. This inconsistency can make the synthesized speech sound less natural. In this paper, we propose a statistical speech synthesis system based on a hidden semi-Markov model (HSMM), which can be viewed as an HMM with explicit state duration PDFs. The use of HSMMs can solve the above inconsistency because we can incorporate the state duration PDFs explicitly into both the synthesis and the training parts of the system. Subjective listening test results show that use of HSMMs improves the reported naturalness of synthesized speech.},
keywords={},
doi={10.1093/ietisy/e90-d.5.825},
ISSN={1745-1361},
month={May},}

Copy

TY - JOUR
TI - A Hidden Semi-Markov Model-Based Speech Synthesis System
T2 - IEICE TRANSACTIONS on Information
SP - 825
EP - 834
AU - Heiga ZEN
AU - Keiichi TOKUDA
AU - Takashi MASUKO
AU - Takao KOBAYASIH
AU - Tadashi KITAMURA
PY - 2007
DO - 10.1093/ietisy/e90-d.5.825
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E90-D
IS - 5
JA - IEICE TRANSACTIONS on Information
Y1 - May 2007
AB - A statistical speech synthesis system based on the hidden Markov model (HMM) was recently proposed. In this system, spectrum, excitation, and duration of speech are modeled simultaneously by context-dependent HMMs, and speech parameter vector sequences are generated from the HMMs themselves. This system defines a speech synthesis problem in a generative model framework and solves it based on the maximum likelihood (ML) criterion. However, there is an inconsistency: although state duration probability density functions (PDFs) are explicitly used in the synthesis part of the system, they have not been incorporated into its training part. This inconsistency can make the synthesized speech sound less natural. In this paper, we propose a statistical speech synthesis system based on a hidden semi-Markov model (HSMM), which can be viewed as an HMM with explicit state duration PDFs. The use of HSMMs can solve the above inconsistency because we can incorporate the state duration PDFs explicitly into both the synthesis and the training parts of the system. Subjective listening test results show that use of HSMMs improves the reported naturalness of synthesized speech.
ER -