Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Identification of Japanese Double-Mora Phonemes Considering Speaking Rate for the Use in CALL Systems

Carlos Toshinori Ishi, Keikichi Hirose, Nobuaki Minematsu

Department of Information and Communication Engineering, University of Tokyo, Japan

Effect of speaking rate on phone duration was investigated to improve a Computer-Aided Language Learning (CALL) system that teaches Japanese double-mora phoneme pronunciation [1]. The influence of the speaking rate was formulated as a linear regression line for each phone, based on the analysis results of a male speakerís non-sense word utterances. The formulations were used to develop a method to estimate the speaking rate of the utterance over two hypothesis (presence and absence of double-mora phoneme), and decide which of the hypothesis is the more suitable. Experiments indicated that good identification between singleand double-mora phonemes was possible irrespective of the speaking rate.


  1. Kawai and Hirose, "Teaching the pronunciation of Japanese double-mora phonemes using speech recognition technology," Speech Communication, Vol.30, Nos.2-3, pp.131-143, (2000).

Full Paper

Bibliographic reference.  Ishi, Carlos Toshinori / Hirose, Keikichi / Minematsu, Nobuaki (2000): "Identification of Japanese double-mora phonemes considering speaking rate for the use in CALL systems", In ICSLP-2000, vol.1, 786-790.