A comprehensive analysis, synthesis system and synthesis methodology for the production of high quality speech

摘要

The paper presents a complete speech synthesis system, which addresses the problems of accurate formant analysis and synthesis to produce high quality natural-sounding multi-gender/multi-language speech. The system comprises two main elements; the Nottingham Parallel Seven Formant Speech Synthesizer implemented on a multi-transputer network for real time execution, and a highly accurate formant analysis system based on the group delay function. The paper also presents a systematic approach to the definition of a speech demisyllable database which can be used for parametric or non-parametric speech synthesis of the German language. The results of the analysis and synthesis of speech sounds and a male German utterance are presented.