Study on automatic prediction of sentential stress for Chinese Putonghua Text-to-Speech system with natural style  

作　　者：SHAO Yanqiu HAN Jiqing ZHAO Yongzhen LIU Ting 

机构地区：[1]School of Computer Science and Technology, Harbin Institute of Technology Harbin 150001

出　　处：《Chinese Journal of Acoustics》2007年第1期49-62,共14页声学学报（英文版）

基　　金：This work was supported by the National Natural Science Foundation of China (No. 60085001)

摘　　要：Stress is an important parameter for prosody processing in speech synthesis. In this paper, we compare the acoustic features of neutral tone syllables and strong stress syllables with moderate stress syllables, including pitch, syllable duration, intensity and pause length after syllable. The relation between duration and pitch, as well as the Third Tone （T3） and pitch are also studied. Three stress prediction models based on ANN, i.e. the acoustic model, the linguistic model and the mixed model, are presented for predicting Chinese sentential stress. The results show that the mixed model performs better than the other two models. In order to solve the problem of the diversity of manual labeling, an evaluation index of support ratio is proposed.Stress is an important parameter for prosody processing in speech synthesis. In this paper, we compare the acoustic features of neutral tone syllables and strong stress syllables with moderate stress syllables, including pitch, syllable duration, intensity and pause length after syllable. The relation between duration and pitch, as well as the Third Tone （T3） and pitch are also studied. Three stress prediction models based on ANN, i.e. the acoustic model, the linguistic model and the mixed model, are presented for predicting Chinese sentential stress. The results show that the mixed model performs better than the other two models. In order to solve the problem of the diversity of manual labeling, an evaluation index of support ratio is proposed.

分 类 号：H102[语言文字—汉语]