Fourth ISCA ITRW on Speech Synthesis

August 29 - September 1, 2001
Perthshire, Scotland

A database design for a concatenative speech synthesis system for the disabled

Akemi Iida (1,2) and Nick Campbell (2,3)

(1) Keio Research Institute at SFC, Keio University, Kanagawa, Japan
(2) Japan Science and Technology Corporation, CREST
(3) Information Sciences Division, ATR International, Kyoto, Japan

This paper reports on our research on designing a speech corpora in Japanese for a concatenative speech synthesis system that is to be used for a specific purpose. For this work the purpose was set to assist communication for non-vocal people. Four kinds of source database for synthesis were developed by combining different speech corpora created from read speech of an Amyotropic Lateral Sclerosis (ALS) patient who was anticipating the imminent loss of his voice. This work confirmed that the recording of a minimum set of phonetically balanced sentences (129 sentences) was insufficient for concatenative speech synthesis and that a combinAion of these and a recording of well-read continuous-text material produced more natural sounding synthesized speech. A communication aid was developed using a concatenated speech synthesis with the database created in this work.

Full Paper

Bibliographic reference.  Iida, Akemi / Campbell, Nick (2001): "A database design for a concatenative speech synthesis system for the disabled", In SSW4-2001, paper 135.