The Study And Application Of Text-to-Speech System

Combined with the national nature and science fund item "Key Technology Study for Chinese Spoken Language Translation" , "Research on Language Processing in Speech-to-Speech Machine Translation" and engineering requirements, the theories and techniques about CTS (Chinese Text-to-Speech) and SST (Speech-to-Speech Translation) are studied in this thesis, and the high quality CTS system and dual (English-Chinese/Chinese-English) SST system are completed. The main contribution of the dissertation is classified into three parts, specifically these are:(1) The overseas CTS techniques are summarized and analyzed. Both linguistic and acoustic processing methods in the CTS system are discussed, and the CTS system is accomplished based on PSOLA (Pitch-Synchronous-Overlap-Add) technique.(2) The CTS and ETS (English Text-to-Speech) software are developed based on the PDA (Personal Data Assistant) platform and the new Chinese and English special text-to-speech chips, which are applicable arbitrary text, are exploited.(3) The overseas SST techniques are summarized, and the dual speech-to-speech translation system named Huajian Speech Translation Test System are completed based on the text-to-speech technique discussed above, the speech recognition technique of IBM ViaVoice and the English-Chinese/Chinese-English machine translation techniques of Huajian.