Modeling of articulatory dynamics using cascaded first-order systems

Abstract

This paper proposes modeling of articulatory dynamics using cascaded first-order systems. The model, which consists of simple first-order components, can be a good approximation for the behavior of a complex system. In the paper, the approach to modeling the articulator movements was performed on several articulatory organs: tongue, jaw, etc. Parameters describing the model were obtained from observed articulatory data in such a way that time patterns of impulse response of the model were approximated to velocity ones of movements. Good approximations were obtained on time patterns of articulation for vowels and consonants. Therefore, successive movement patterns for continuous speech including consonants were approximated by combining several models with high accuracy. By modeling articulatory dynamics for vowels and consonants based on articulatory measured data, complex articulatory patterns in a form of [-V C V-] were synthesized and the time instant of input commands for articulation of vowels and consonants were estimated in a connected speech sound including consonants. [Work supported by SCAT.]