A Mandarin TTS system with an integrated prosodic model

Author

Pin, ShaoHuang ; Lee, Yehlin ; Chen, Yong-Cheng ; Wang, Hsin-Min ; Tseng, Chiu-Yu

Author_Institution

Phonetics Lab, Acad. Sinica, Taipei, Taiwan

fYear

2004

fDate

15-18 Dec. 2004

Firstpage

169

Lastpage

172

Abstract

Phrase grouping is essential to characterize the prosody of Mandarin fluent speech. Evidence of prosodic phrase grouping has been found both in adjustments of F₀ contours and temporal allocations within and across phrases. We discuss the development of a Mandarin TTS system that integrates prosody processing modules, such as duration modeling, F₀ modeling, and break predictions. The database consists of 1292×3 syllable-tokens chopped off specially designed three-phrase carrier sentences. Since temporal allocations and rhythmic structure in speech flow are carefully dealt with, the TTS system is capable of converting long paragraph text input into natural synthesized speech output.

Keywords

natural language interfaces; speech synthesis; text analysis; Mandarin TTS system; break predictions; duration modeling; fluent speech; integrated prosodic model; long paragraph text; natural synthesized speech; prosodic phrase grouping; rhythmic structure; temporal allocations; three-phrase carrier sentences; Bismuth; Databases; Humans; Information science; Labeling; Modems; Natural languages; Predictive models; Speech analysis; Speech synthesis;

fLanguage

English

Publisher

ieee

Conference_Titel

Chinese Spoken Language Processing, 2004 International Symposium on

Print_ISBN

0-7803-8678-7

Type

conf

DOI

10.1109/CHINSL.2004.1409613

Filename

1409613