4th International Conference on Spoken Language Processing

Philadelphia, PA, USA
October 3-6, 1996

An Improved Vocal Tract Model of Vowel Production Implementing Piriform Resonance and Transvelar Nasal Coupling

Jianwu Dang (1), Kiyoshi Honda (2)

(1) ATR Human Information Processing Res. Labs., Kyoto, Japan
(2) University of Wisconsin, Madison, WI, USA

This paper proposes an improved vocal tract model of vowel production, which incorporates acoustic effects of the piriform fossa and transvelar nasal coupling. In this study, the vocal tract model was derived from the MRI data of a subject. The piriform fossa was modeled based on the MRI data as a side branch of the vocal tract. The velum wall was modeled as a cascaded impedance of a viscous resistance, a mass and a stiffness, which were estimated by acoustic and mechanical experiments conducted on three subjects. Transfer functions of vowels /a/ and /i/ were computed under the conditions of with and without the piriform fossa and transvelar nasal coupling. The results showed that both the piriform fossa and transvelar coupling play important roles in shaping the first two formants for closed vowels, while the piriform fossa is the main factor affecting the formants of open vowels. By comparing computed transfer functions with real speech spectra for the same subject, it is clarified that our improved model gives a more realistic performance than the traditional model.

Full Paper

Bibliographic reference.  Dang, Jianwu / Honda, Kiyoshi (1996): "An improved vocal tract model of vowel production implementing piriform resonance and transvelar nasal coupling", In ICSLP-1996, 965-968.