First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

English Speech Training Using Voice Conversion

Keiko Nagano, Kazunori Ozawa

C&C Information Technology Research Labs., NEC Corporation, Kanagawa, Japan

This paper proposes an English prosody training method using voice conversion technique. The unique point of this proposed method is that voice converted synthetic speech is used to train English prosody pronunciation. The synthetic speech is produced by converting important prosodic parameters in the student's speech into corresponding native English speaker's speech, while the student's voice characteristics except prosody are preserved. By using the proposed method, pronunciation problem can be easily found out, and training efficiency is improved. The comparative evaluations for training efficiency of the proposed training method and the conventional method show that the proposed method, using the voice converted synthetic speech, is more effective in English prosody training than the conventional method, using the native English speaker's original speech.

Full Paper

Bibliographic reference.  Nagano, Keiko / Ozawa, Kazunori (1990): "English speech training using voice conversion", In ICSLP-1990, 1169-1172.