An interactive construction system of three-dimensional vocal tract shapes from tomogram

Abstract

A new interactive system is developed to construct the 3-D vocal tract shape from tomogram, for the purpose of understanding the relationship between the speech production process and the geometrical figure of the human vocal tract. The shaded 3-D vocal tract shape can be seen with this system from any direction, by construction from a set of 2-D images measured by, for example, MRI or x-ray CT, and basic geometrical information is obtained for estimating the area function and transfer function, or creating the mesh model for 3-D FEM. For constructing a 3-D vocal tract shape, the following three operations are conducted for the system. First, the 3-D locations of the tomogram images are interactively determined. Second, the contours of the air tract in the images are extracted after several image processings. Third, the wire frame model of the vocal tract, including a branching case, is constructed computationally from the extracted contours. The system is developed using the C++ language as an object-oriented programming language, Motif toolkit as GUI, and PEX as 3-D graphic libraries. The proposed system provides a convenient tool for speech research on the X-Window environment.