A novel 3D geometric articulatory model

This study employs a novel method to construct a geometric model based on Magnetic Resonance Imaging data by taking the physiological boundaries of speech apparatus into account. Two improvements are made in the modeling process: i) it combines the images of different views to help the outline annotation process; ii) it models speech organs´ meshes with reference to the anatomical structures. Both qualitative and quantitative evaluations indicate that the proposed method surpasses the conventional method. Based on the meshes of the speech organs for different articulations, the linear component analysis is implemented to extract the control parameters. Each speech organ can be described by no more than three control parameters. After the reconstruction, the average error between model and real data is less than 1.0 mm. This is also the first effort to construct a 3D vocal tract model based on Chinese MRI data. It will facilitate the theoretical study and practical applications for Chinese speech production related issues.