Shi, Bo; (1990) A speech synthesis-by-rule system for Modern Standard Chinese. Doctoral thesis (Ph.D), UCL (University College London).

Text A_speech_synthesis-by-rule_sys.pdf Download (18MB)

Abstract

This thesis discusses the development of a Chinese speech synthesis-by-rule system and presents the structure and features of the system. The aim of the work is to produce highly intelligible standard Chinese speech with natural sounding intonation from unrestricted Chinese text, using a parallel formant speech synthesizer. The synthesis system accepts standard Chinese Pinyin text as input, either from a conventional keyboard or from a computer readable file. Text to detailed phonetic description conversion is carried out in three steps: 1) application of a group of phonological and phonetic rules to convert Pinyin text into demisyllable strings; 2) conversion of the demisyllables into a succession of phonetic elements using a dictionary look-up strategy; 3) application of prosodic rules at different levels. Two specific features of this text-to-phonetic conversion system are: the use of a specially designed demisyllable dictionary which permits an effective and reliable way of deriving phonetic elements from text, and the implementation of a new generative intonation framework which enables a very wide variety of natural sounding fundamental frequency contours to be generated automatically. The acoustic-phonetic rules used in the system are developed from the Holmes-Mattingly-Shearme (HMS) algorithm. A complete set of Chinese phonetic tables, each of which contains the acoustic properties and co-articulation information of an acoustic segment, has been developed on the basis of systematic acoustic-phonetic analysis of standard Chinese syllables. The synthesis system has been implemented on a BBC microcomputer to drive a LSI parallel formant synthesizer. An IBM PC version of the system has also been developed. A segmental intelligibility test shows that the performance of the present system is quite comparable with those of the best English speech synthesis-by-rule systems now available.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item