Abstract:

The purpose of this study is to incorporate accurate muscle morphology in a computational 3-D tongue model based on anatomical literature and MRI data. Previous work [C.-M. Wu and R. Wilhelms-Tricarico, J. Acoust. Soc. Am. 100, 2659(A) (1996)] defined several landmarks within the 3-D space of the tongue tissue and developed a computational method to map Miyawaki's drawing of muscle fiber directions [K. Miyawaki, Ann. Bul. RILP, Univ. Tokyo 8, 23--50 (1974)] onto the cross-sectional tongue images from the visible human data. In this study, the same technique is applied to identify a set of corresponding landmarks between Miyawaki's data and MRI scans of a subject's tongue forming a consistent articulatory posture. The geometrical mapping of these two data sets is achieved by the thin-plate spline mapping method. The result shows the sufficiency and robustness of the chosen landmark points to represent a general 3-D structure of the human tongue. This study is expected to provide a computational basis for quantitative comparison of MRI data across subjects and vowel types.