Synthesizing Japanese intonation using a downstep model

Publisher Website
Google Scholar
Abstract
Studies of Japanese intonation have unanimously shown that fundamental frequency (F0) tends to fall during the course of an utterance. Previous synthesis systems for Japanese have all modeled this F0 trend using the single mechanism of a declining overall tonal reference line that is reset at the beginning of each utterance or major phrase. Recent research by Poser, however, suggests that a more accurate model must account for most of the F0 decline by means of “downstep,” a sudden step-like reduction of the reference line that is triggered by the accentual pattern of the preceding minor phrase [W. J. Poser, “On the mechanism of F0 downdrift in Japanese,” J. Acoust Soc. Am. Suppl. 1 74, S89 (1983)]. This paper describes a synthesis program for Japanese intonation that incorporates downstep. The program is being used to investigate the phonetic correlates of intonational phrasing and the interactions between syntactic phrasing and intonation. A demonstration tape will be played to illustrate the synthesis program.

This publication has 0 references indexed in Scilit: