Acoustic analysis of the vocal tract during vowel production by finite-difference time-domain method

Abstract

An acoustic simulator based on the finite-difference time-domain (FDTD) method was evaluated by acoustic measurements on solid models of the vocal tract. Three-dimensional vocal tract (3D VT) shapes for a male subject during production of the five Japanese vowels were measured by magnetic resonance imaging. Transfer functions of the 3D VT shapes were computed by the acoustic simulator. The accuracy of the finite-difference algorithm was second-order in time and fourth-order in space. From the same 3D VT shapes, solid models were made of epoxide resin by a stereo-lithographic technique, and their acoustic transfer functions were measured using a time-stretched pulse signal. The calculated and measured spectra were compared up to 8 kHz. Although locations of major poles and zeros were common between calculated and measured spectra, the calculated spectra showed a larger number of small zeros, possibly caused by underestimation of VT acoustic losses in the simulation. However, the lower four formants were simulated remarkably accurately, with a mean absolute error of only 2.2 % compared with the acoustic measurements. These facts indicated that the acoustic simulator can reasonably account for acoustic phenomena within the VT.

One-dimensional and three-dimensional propagation analyses of acoustic characteristics of Japanese and French vowel /a/ with nasal coupling

“…2 with area functions). The pole/zero pair at approximately 3,300 Hz can be ascribed to an interdental space [18]. The formant at approximately 4,400 Hz is probably an oral formant shifted by the presence of a pole/ zero pair at approximately 4,700 Hz owing to the piriform sinuses, which are not modeled for the French subject.…”

Finite element analysis of fine-structural effect on acoustic characteristics of vocal tract

“…Dang and Honda reported that the piriform fossa contributes strong troughs to the VTTFs, as determined from measurement experiments using an MRIbased mechanical model [2]. Takemoto et al reported the acoustical effects of the piriform fossae, epiglottic valleculae, and interdental spaces determined from measurement experiments and computer simulation using MRI-based physical models and finite-difference time-domain analysis data [3]. However, when the model is geometrically smoothed, the extra poles and zeros sometimes disappear in the VTTFs.…”

Synthesis of VV Utterances from Muscle Activation to Sound with a 3D Model

“…The latter boundary condition does not consider radiation losses. These could be implemented by extending the computational domain outside the vocal tract [20,21], but at the prize of increasing the computational cost. The sequence [Ai] was generated by numerically solving the ALE mixed wave equation (1) with boundary and initial conditions (2) using the Finite Element Method (FEM).…”