INTERSPEECH 2008
9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Estimation of Vocal Tract Area Function for Mandarin Vowel Sequences Using MRI

Gaowu Wang (1), Jianwu Dang (1), Jiangping Kong (2)

(1) JAIST, Japan; (2) Peking University, China

To fully explore the dynamic properties of speech production and investigate the relation between vocal tract geometry and speech acoustics, estimation of vocal tract area functions from measurements of the sagittal plane is an important step. In this study, we investigated the relation between the measurements on two dimensional (2D) and three dimensional (3D) MRI data and used an alpha-beta model to describe this relation. As a result, a set of parameters were derived from 3D static MRI data, and applied to time-varying vocal tract widths derived from 2D MRI movies, to synthesize Mandarin vowel sequences. An acoustic evaluation comparing the natural and calculated formants shows that the alpha-beta model can represent dynamic states of articulatory movements of vowel sequences, as well as those of the sustained vowels.

Full Paper

Bibliographic reference.  Wang, Gaowu / Dang, Jianwu / Kong, Jiangping (2008): "Estimation of vocal tract area function for Mandarin vowel sequences using MRI", In INTERSPEECH-2008, 1182-1185.