Skip to content

Detail of publication

Citation

Tihelka, D. and Matoušek, J. : Revealing the most significant deterioration factors in single candidate synthetic speech . Specom 2005, proceedings of 10th International Conference SPEECH and COMPUTER, p. 171-174, Moscow State Linguistic University, Moscow , 2005.

Abstract

The paper focuses on revealing the factors that cause the deterioration of the naturalness of synthetic speech. Modified listening tests are used for this task, as we also need to determine the impact of each considered factor. Synthetic speech generated by the single candidate version of our TTS was examined with the use of modified listening tests, and the most prominent deterioration factors were evaluated on two corpora. The most significant factor turned out to be the extensive modification of units, followed by segmentation inaccuracy and spectral discontinuities. However, the last two were not so significant. Listening tests described in this paper can be used not only for testing single candidate TTS systems, but also for testing other types of TTS. In fact, what is required to test depends only on the definition of the set of factors.

Detail of publication

Title: Revealing the most significant deterioration factors in single candidate synthetic speech
Author: Tihelka, D. ; Matoušek, J.
Language: English
Date of publication: 17 Oct 2005
Year: 2005
Type of publication: Papers in proceedings of reviewed conferences
Book title: Specom 2005, proceedings of 10th International Conference SPEECH and COMPUTER
Page: 171 - 174
ISBN: 5-7452-0110-X
Publisher: Moscow State Linguistic University
Address: Moscow
Date: 17 Oct 2005 - 19 Oct 2005
/ 2008-05-19 15:29:56 /

Keywords

syntetic speech quality, deterioration factors, single instance TTS, extensive modification, segmentation inaccuracy, spectral discontinuities

BibTeX

@INPROCEEDINGS{TihelkaD_2005_Revealingthemost,
 author = {Tihelka, D. and Matou\v{s}ek, J.},
 title = {Revealing the most significant deterioration factors in single candidate synthetic speech},
 year = {2005},
 publisher = {Moscow State Linguistic University},
 address = {Moscow },
 pages = {171-174},
 booktitle = {Specom 2005, proceedings of 10th International Conference SPEECH and COMPUTER},
 ISBN = {5-7452-0110-X},
 url = {http://www.kky.zcu.cz/en/publications/TihelkaD_2005_Revealingthemost},
}