Second ESCA/IEEE Workshop on Speech Synthesis

September 12-15, 1994
Mohonk Mountain House, New Paltz, NY, USA

A 3-D Model of the Lips for Visual Speech Synthesis

Thierry Guiard-Marigny, Ali Adjoudani, Christian Benoît

Institut de la Communication Parlée, INPG/ENSERG, Université Stendhal, Grenoble, France

Unlike most of the regions of the human face, lips are essentially characterized by their border contours. The internal and external contours of the vermilion zone can be fitted by means of algebraic equations. The coefficients of these equations must be controlled so that the lip shape can be adapted to various speakers conformations and to any speech gesture. To reach this goal, the 3-D model of the lips here described has been worked out from geometrical analysis of the natural lips of a French speaker videotaped when uttering the most representative coarticulated strings of French phonemes. The reference labial database we used was made of 22 lip-jaw shapes that constitute the "labial space" of a French speaker and of the most relevant parameters. From this, a 2-D lip model was developed to adjust a set of continuous functions that best fit the front contours of the 22 "visemes". Then all the various equation coefficients were predicted from only three anatomical parameters which can easy to measure on the speaker's face. This model was then extended to 3D. Equations of the lip contours in the axial plane was similarly obtained. Volume was then given to the lips by linearly interpolating three intermediate contours in between the internal and external ones. Ultimately, five parameters are necessary to predict all the equations of this 3-D model.

Full Paper

Bibliographic reference.  Guiard-Marigny, Thierry / Adjoudani, Ali / Benoît, Christian (1994): "A 3-d model of the lips for visual speech synthesis", In SSW2-1994, 49-52.