Chinese is a tone language. For a tone, the characters of its F0 pitch contours will be quite different in the condition of continuant speaking from the isolated speaking. The present researches about the Chinese tone are still centralized on the isolated speaking one, and about tone in fluent speech, there are some statements about the phenomenon of the two-word, threeword, four-word co-reading, but it can not represent the real characters of the tone in fluent speech very well.
Our work bases on the spoken Chinese. And with the methods of statistics and classifications, we study and analyze the characters and behaviors of the F0 pitch contours of Chinese mandarin four tones in fluent speech.
In the Paper, we put forward a mathematical model for describing the pitch contours: pitch contours time normal model (PiCTN model). In Chinese fluent speech, because of the coreading and different surroundings, the pitch contours of one tone will be quite different from what it will be in the isolated speaking, and will even have many different characters. And about the dynamic characters of pitch contours for Chinese Putonghua four tones in fluent speech, we also give the conclusion: the F0 pitch contour of tone 1 should be high and flat, and sometimes it will be rising with the even speed, the contour of tone 2 will be rising with the accelerated speed, and the pitch contour of tone 3 will drop with the decreasing speed, and lastly the F0 contour of tone 4 should be dropping with the speed increasing.
The results of our work will be useful for the prosody modeling in Text-to-Speech (TTS) systems and the tone confirmation in the speech recognition.
Cite as: Wu, Z., Cai, L., Zhou, T. (2000) Research on dynamic characters of Chinese pitch contours. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 3, 686-689, doi: 10.21437/ICSLP.2000-628
@inproceedings{wu00f_icslp,
author={Zhiyong Wu and Lianhong Cai and Tongchun Zhou},
title={{Research on dynamic characters of Chinese pitch contours}},
year=2000,
booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
pages={vol. 3, 686-689},
doi={10.21437/ICSLP.2000-628}
}