Tongue Visualization Model for Mandarin Pronunciation Based on MRI

An improved tongue statistical model was presented. Volume images of tongue were acquired by MRI from one subject uttering a corpus representative of Chinese vowel-consonant articulations. Three-dimensional tongue surface outline was represented by a mesh obtained by the set of tongue contours traced from the volume images. The set of the 3D coordinates of all vertices of the tongue mesh constituted the variables on which linear component analysis was applied and subsequently six linearly independent components which has physiological significance were firstly extracted. Jaw parameter, the first of these six parameters, can control tongue rotation. Parameters of tongue body and tongue dorsal control respectively tongue forward and backward, flattening and bunching movements, and the last three tongue tip parameters control tongue tip up and down, anterior and posterior, and upward movements respectively. These six parameters can explain 87.4% of the variance of the tongue data, which has a good simulation effect and better than other research. The method proposed reduced the complexity of tongue modeling and can be effectively applied to the modeling of tongue for Chinese Mandarin pronunciation.