The study of prosodic adjustment in Chinese speech synthesis.

Abstract

In the recent study of Chinese speech synthesis, more attention is focused on improving the naturalness of the synthetic speech. The key point is how to design a flexible speech synthesizer, which can modify the prosodic features effectively. As a platform, a Chinese text-to-speech (C-TTS) system with a large word database is used for our experiments. The principle of prosodic adjustment in these experiments is based on source/filter theory and the homomorphic processing method is used. A syllable parameter database and a prosodic model are constructed carefully to implement Chinese prosodic adjustment. The result of the experiments shows that this method is better than the TD-PSOLA method. A PCplugboard based on a DSP chip “TMS320C54X” is being used to speed up the computation power, so it can be expected that a real time C-TTS system with prosodic adjustment will be successfully developed.