Study of prosody model on Chinese speech synthesis based on the classification of syllabic prosody features

A prosody modeling method based on statistic model is described. Based on this, a Chinese prosody model based on the classification of syllabic prosody features is presented, which makes automatic prosody prediction with prosody templates and prosody cost function. And the automatic training algorithm of the model in detail is described. Further more, according to statistic prosody modeling method, the influence to prosody template selection with the help of the analysis of the prosody interaction among prosody elements is analyzed. Finally, the error distribution of the statistic method based prosody prediction is given. The results show good naturalness and much flexible in application.