Tvorba MKP modelů vokálního traktu pro české samohlásky

FEM, vocal tract, czech vowel, MR, meshing, modal analysis

This article is dealing with problems of creating FEM models of human vocal tract of Czech vowels. The solution is connected to grant project GA 106/98/K019. We need the FEM models for modal analysis of vocal tract. To take geometry of vocal tract we have scanned some person on MR. Scanning was in time of speaking of specified vowels. First step of creating FEM models was segmentation of vocal tract from MR data. Result of segmentation are voxel models. Next was vectorization of the voxel models by Marching cubes method, witch produce boundary of the voxel models, as a triangular meshes. Next the meshes was smoothed and decimated. Then we use them as a base for Delaunay triangulation. Results of the one are tetrahedral meshes. Last step is optimizing quality of the tetrahedral meshes, to take final FEM models of vocal tract.