The speech synthesis system for an unlimited Japanese vocabulary

This paper describes the speech synthesis system for an unlimited Japanese word. In the system, we adopt a new time domain synthesis method using symmetric waveforms for improvement of synthetic speech quality; featuring that the pitch is well controlled, the synthetic speech is very clear, and the synthetic algorithm is simple. In addition, owing to the information-compression processing such as waveform symmetrizing and ADPCM encoding, the required speech data capacity is relatively low, namely, 60k bytes. This is fully in practical range. As the result of performing the evaluation test on the synthetic speech with the system, the following data are obtained: The sound articulation score of CV-syllables: 77% The intelligibility score of words: 99.5%