Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Computation of 3-D Vocal Tract Acoustics Based on Mode-Matching Technique

Kunitoshi Motoki (1), Xavier Pelorson (2), Pierre Badin (2), Hiroki Matsuzaki (1)

(1) Department of Electronics and Information Engineering, Hokkai-Gakuen University Sapporo, Japan
(2) Institut de la Communication Parlée, UMR CNRS 5009, INPG-Université Stendhal, Grenoble, France

A model for calculating the acoustic characteristics of 3-D vocal tract configuration is presented. A mode-matching technique and an impedance transformation are used to calculate the higher-order modes in this model where both propagative and evanescent higher-order modes are considered. A cascaded structure of rectangular acoustic tubes is introduced as an approximation of the real vocal tract geometry. The number of higher-order modes in each tube can be selected independently, which significantly decreases the instability of computation caused by the evanescent higher-order modes. Preliminary calculation results are discussed from the view point of the influences of the offset connection of tubes.


Full Paper

Bibliographic reference.  Motoki, Kunitoshi / Pelorson, Xavier / Badin, Pierre / Matsuzaki, Hiroki (2000): "Computation of 3-d vocal tract acoustics based on mode-matching technique", In ICSLP-2000, vol.1, 461-464.