EIGENMODES IN SPEECH DYNAMICS

Project: Research project

Description

The long-term goal of this research is to investigate and model possible strategies that a speaker might employ controlling the acoustic eigenmodes (formants) of the vocal tract during dynamic speech. A second set of eigenmodes, which are geometric rather than acoustic. The geometric eigenmodes are considerable to be a limited set of canonical movement patterns, representing coordinated actions of the tongue, jaw, lips, velum, and larynx. Dynamic speech is assumed to be carried out by superimposing the geometric eigenmodes on a neutral vocal tract shape. It is further hypothesized that the shape of the geometric eigenmodes is at least similar across speakers; and that the speaker-specific vocal tract characteristics are embedded in the neutral state of speech airway system. Eigenmodes are assumed to be recoverable from either static or dynamic image data. To test these hypotheses the following specific aims are proposed: (1) to develop area function and midsagittal width inventories of vocal tract shapes using magnetic resonance imaging (MRI) on each of six subjects; (2) to develop a kinetic vocal tract model based on geometric eigenmodes; (3) to develop a strategy for kinematic vocal tract control of formant trajectories based on recorded vowel-vowel transitions, vowel-consonant-vowel utterances, and sentences; (4) to study the vocal tract scaling factors between males, females, and children; (5) to interpret midsagittal speech movement, observable. From x-ray microbeam data, in terms of geometric eigenmodes; and (6) to conduct perceptual experiments that assess the intelligibility of the speech stimulated by the kinematic vocal tract model. The outcome of this research will extend our current knowledge about the relationship between articulation and acoustics, particularly with regard to dynamic aspects. The ideas developed may have eventual application to the development of speech synthesis and recognition systems, for understanding articulation disorders, and possibly influencing speech training and rehabilitation programs.
StatusFinished
Effective start/end date9/1/0012/31/12

Funding

  • National Institutes of Health: $244,318.00
  • National Institutes of Health: $246,015.00
  • National Institutes of Health: $249,256.00
  • National Institutes of Health: $221,808.00
  • National Institutes of Health: $283,891.00
  • National Institutes of Health: $246,015.00
  • National Institutes of Health: $256,629.00
  • National Institutes of Health: $240,848.00
  • National Institutes of Health: $243,554.00
  • National Institutes of Health: $235,016.00

Fingerprint

Kinematics
Speech intelligibility
Acoustics
Experiments
Speech synthesis
Magnetic resonance
Speech recognition
Patient rehabilitation
Speech analysis
Trajectories
Imaging techniques
X rays
Kinetics

ASJC

  • Medicine(all)
  • Neuroscience(all)