The Synthesis Rules in a Chinese Text-to-Speech System
Journal
IEEE Transactions on Acoustics, Speech, and Signal Processing
Journal Volume
37
Journal Issue
9
Pages
1309-1320
Date Issued
1989
Author(s)
Abstract
Abstract-An initial attempt to develop a preliminary Chinese text-to-speech system has been made recently. The design approaches are based on a syllable concatenation concept due to the special character-istics of the syllabically paced nature of Chinese language. This paper describes in some detail the synthesis rules developed for this system with special attention given to the lexical tones and other prosodic rules such as concatenation rules, sandhi rules, stress rules, intonation pat-terns, syllable duration rules, pause insertion rules, and energy mod-ification rules. These rules were derived basically from the acoustic properties of Mandarin Chinese, and therefore are useful not only in designing other Chinese text-to-speech systems, but in understanding the characteristics of Mandarin sentences and processing Mandarin speech signals for other purposes such as segmentation or recognition. I.
Type
journal article