Kazuya Takeda,Yoshinori Sagisaka and Hisao Kuwabara
On sentence level factors gaverning segmental duration
in Japanese
Abstract:In this paper, the durational characteristics of Japanese are statistically analyzed
aiming at establishing a fine duration setting rule. First, statistic duration
control model is formulated through factor analysis on a large scale isolated word
database(5,240 words). Second, using 282 continuous sentence speech, each
segmental duration, calculated by this model, is compared with taht of measured
one. The error analysis clarified the existence of the following sentence level
duration control: 1) Pre-pausal lengthening is greater than simple word final
lengthening, 2) Shorter durations are found at sentence final position in
declarative sentences. By applying those sentence level controls to the duration
setting rule for sentence speech, it was turned out that the prediction error can be
reduced to the same amount of errors in isolated speech.