5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

Reconstructing the Tongue Surface from Six Cross-Sectional Contours: Ultrasound Data

Andrew J. Lundberg (1), Maureen Stone (2)

(1) Johns Hopkins University, Department of Computer Science, USA
(2) University of Maryland Medical School, Division of Otolaryngology, USA

This paper presents a method for reconstructing 3D tongue surfaces during speech from 6 cross-sectional contours. The method reduces the dimensionality of the tongue surface and maintains highly accurate reproduction of local deformation features. This modification is an essential step if multi-plane tongue movements are to be reconstructed practically into tongue surface movements. Six cross-sectional contours were used to reconstruct 3D tongue surfaces and these were compared to reconstructions from 60 contours. The best set of 6 cross-sectional contours was determined from an optimized set of 6 midsagittal points. These points had been optimized to predict the midsagittal contour. Errors and reconstruction coverage for the midsagittal optimization were comparable to those resulting from an optimization over the entire surface, indicating this was an adequate method for calculating a sparse data set for use in reconstructing 3D tongue surface behavior.

Full Paper

Bibliographic reference.  Lundberg, Andrew J. / Stone, Maureen (1998): "Reconstructing the tongue surface from six cross-sectional contours: ultrasound data", In ICSLP-1998, paper 0865.