Acoustic characteristics of solid vocal tracts modeled from ATR MRI database of Japanese vowel production

Abstract

"ATR MRI database of Japanese vowel production" provides volumetric magnetic resonance images and speech data of the five Japanese vowels produced by a male native Japanese speaker. In this study, the database was used to evaluate acoustic characteristics of vocal tracts for five Japanese vowels; we measured frequency responses of realistic vocal tract solid models formed by a stereo-lithographic technique. The model's glottis was sealed with a plastic plane with a 1.2-mm hole. A time-stretched pulse signal generated from a horn driver unit was introduced into the solid model at the lip end. The response signals of the models were recorded at the model's glottis from the hole by a probe microphone. This method permits accurate measurement of acoustic characteristics of the vocal tract including the laryngeal cavity, which generates the laryngeal cavity resonance during closed-glottis periods of phonation [H. Takemoto et al., JASA, 120, 2228-2238 (2006), T. Kitamura et al., JASA, 120, 2239-2249 (2006)]. The results provide a benchmark for evaluating numerical analysis methods, such as transmission line models, finite element methods, and finite difference-time domain-methods, which have been used to study vocal tract acoustics. [Work supported by SCOPE (071705001) of Ministry of Internal Affairs and Communications, Japan.]