3D finite element analysis of Japanese vowels in elliptic sound tube model

Abstract

Based on vocal tract cross-section data formed by magnetic resonance images of the head during pronunciation of the Japanese vowels /a/, /i/, and /u/, a three-dimensional sound tube model is constructed with its cross section approximated by an equivalent ellipse with a peripheral length of the vocal tract cross section and with the same cross-sectional area. By means of the finite element method, the sound pressure and the particle velocity in the sound tube are computed. For the sound tube model, a vocal tract wall impedance modeling a soft wall is given as the boundary condition. A radiation space to represent the three-dimensional radiation at the radiation lips is attached at the radiation end. From the simulation results, the radiation impedance and the vocal tract transfer function are computed and the results are compared with the conventional one-dimensional model. The first and second formants agree well with those given by the one-dimensional model, whereas there is a difference in the bandwidth and the formant frequency at the third and higher formant frequencies due to higher-order modes in the sound tube. It is also shown that the radiation impedance is more dependent on the shape of the radiation aperture than on the shape of the sound tube and the wall boundary conditions. © 1999 Scripta Technica, Electron Comm Jpn Pt 3, 83(4): 43–51, 2000