Dynamic Articulatory Model of Speech Production Using Computer Simulation
Abstract
A dynamic articulatory model of speech production is described. From a phonemic input the model generates a description of the configuration of the articulatory mechanism in the midsagittal plane. Positions, shapes, velocities, and other descriptive features of the modeled vocal mechanism are contained in the "state" of the model. "Operators" act as agents for modifying the state by trying to manipulate aspects of the state toward abstract "goals" which are associated with phonemes. Goals are only changed discretely in time, and in this way the desired transformation from a discrete phonemic input to a continuous articulatory output is accomplished. The operator-state bifurcation of the model allows some of the natural constraints of the real vocal mechanism to be included similarly in the model. The model exhibits coarticulation effects attributable to phonemes preceding the "current" phoneme since the state configurative position responds only slowly to the goal directed operators owing to physical and physiological limitations. Coarticulation effects attributable to following or future phonemes result from a "look ahead" procedure that may invoke goals of future phonemes when such goals do not conflict with the goals of the current or more immediate phonemes. Thus anticipatory coarticulation results from a mechanism at a higher level than the sluggish response which causes post coarticulation. The repertoire of speech sound types in the present model includes only vowels and stops, but it is felt that the general methodologies are applicable to all speech sounds.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1966
- Accession Number
- ADA459565
Entities
People
- William L. Henke
Organizations
- Massachusetts Institute of Technology