Third International Conference on Spoken Language Processing (ICSLP 94)
A articulatory model which combines laryngeal and supra-laryngeal articulators has been designed based on our morphological and physiological studies of speech production. The model consists of a finite element model of the tongue and a mass-spring model of rigid structures. Static postures of the model are determined by computing equilibrium of muscle forces on all of the components, and are used to calculate cross-sectional areas of the vocal tract and transfer function of the model with reference to volumetric MRI data. The cricothyroid angle determines parameters for vibration of the two-mass model, and the source sounds are coupled to the model to output synthesized vowel sounds. In this model, biomechanical connections among the jaw, the hyoid bone, the tongue and the laryngeal cartilage are represented by mass-spring actions. The so-called tongue-larynx interaction which is observed in natural speech are demonstrated in the simulation. The overall performance of the model during vowels with various F0 levels has been tested by comparison of the acoustic data from the model's output with recorded speech.
Bibliographic reference. Honda, Kiyoshi / Hirai, Hiroyuki / Dang, Jianwu (1994): "A physiological model of speech production and the implication of tongue-larynx interaction", In ICSLP-1994, 175-178.