Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

A Dynamical Articulatory Model Using Potential Task Representation

Masaaki Honda, Tokihiko Kaburagi

Information Science Research Laboratory, NTT Basic Research Laboratories, Kanagawa, Japan

A computational model for converting a phoneme specific task sequence into articulatory movements is presented. The task is defined as the vocal tract gestures in a coordinative structure of an articulatory system and is represented phoneme-specifically in a metric form of weighted squared distance from the task target value, where the magnitude of the weights controls the significance level of each gesture. Articulatory trajectories are determined so as to minimize the objective function represented by the weighted sum of the smoothness and the task metric criteria on the trajectories. We also describe how to produce a variable pattern of coarticulatory behavior from the phoneme-specific task sequence using this model. The computed articulatory trajectories are compared with the empirical data obtained with an electro-magnetic instrument.

Full Paper

Bibliographic reference.  Honda, Masaaki / Kaburagi, Tokihiko (1994): "A dynamical articulatory model using potential task representation", In ICSLP-1994, 179-182.