ON-LINE VISUALIZATION OF SPEECH ORGANS USING MRI: A 3D APPROACH TO SPEECH ARTICULATION MODELING

摘要

A three-dimensional on-line visualization of the vocal tract during speech production was performed based on MRI data obtained from a female speaker producing the six Russian vowels. These images were collected using original method of 3D MRI-scanning where the starting moments of MRI processes enabled co-operative activities from a patient's side via a special remote-control device. A stroboscopic method of data acquisition was used to reconstruct real articulatory processes for each speech stimulus. Data show clear real-time movements of the lips, tongue, underjaw and mandible, as well as velum and facial surfaces. Thus collected animated data could be exposed for improved teaching and learning foreign languages' (in our case, the Russian language) technology, as well as for speech synthesis based on a physiologically relevant articulatory model. Sample movies and data analysis strategies are presented.