Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Discriminating Chinese Lexical Tones by Anchoring F0 Features

Jin-Song Zhang (1), Satoshi Nakamure (1), Keikichi Hirose (2)

(1) ATR Spoken Language Translation Research Labs, Kyoto, Japan
(2) Dept. of Frontier Informatics, Univ. of Tokyo, Japan

In this paper we present some new features, called as anchoring FO features including left-context-dependent and right-context-dependent ones, for discriminating Chinese lexical tones of continuous speech. The features are calculated based on a hypothesis that auditory tone perception possibly depends on the shift of the judgement boundary of tone targets by the anchoring effects from the neighboring tone targets. Statistical distribution analyses on continuous speech data by 20 speakers (10 males and 10 females) showed that the proposed anchoring FO features are statistically efficient for tone discrimination, and the left-context-dependent FO anchoring feature is more efficient than the FO height for discriminating the pitch value of tone onset. Moreover, higher efficiency of the left-context-dependent anchoring features than the right-context-dependent ones indicates that carry-over effect play a more important role than the anticipation effect for tone discrimination.


Full Paper

Bibliographic reference.  Zhang, Jin-Song / Nakamure, Satoshi / Hirose, Keikichi (2000): "Discriminating Chinese lexical tones by anchoring F0 features", In ICSLP-2000, vol.2, 87-90.