Implementation of waveform files for synthesis-by-rule

Author

Hayashi, Takashi ; Murakami, Ken´ya

Author_Institution

NTT DATA Communication Systems Corp., Kanagawa, Japan

Volume

2

fYear

1992

fDate

23-26 Mar 1992

Firstpage

41

Abstract

A synthesis-by-rule method using phoneme-length waveforms is investigated, aimed at improving the naturalness of synthesized speech. A method for implementing waveform files to store synthesis units is proposed. By adopting this method, word-level accent expressions can be handled with proper waveform selection, thus simplifying the traditional prosodic control model. The method´s effectiveness is verified by examining the relationship between accent existence and acoustic parameters for Japanese words with five vowels spoken separately, and by performing a preliminary listening test of short Japanese homonyms

Keywords

speech synthesis; Japanese words; acoustic parameters; implementation; naturalness; phoneme-length waveforms; prosodic control model; short Japanese homonyms; synthesis-by-rule; synthesized speech; vowels; waveform files; waveform selection; word-level accent expressions; Control system synthesis; Data communication; Dictionaries; Frequency; Linear predictive coding; Performance evaluation; Speech coding; Speech processing; Speech synthesis; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1992. ICASSP-92., 1992 IEEE International Conference on

Conference_Location

San Francisco, CA

ISSN

1520-6149

Print_ISBN

0-7803-0532-9

Type

conf

DOI

10.1109/ICASSP.1992.226125

Filename

226125