| |
Abstract:
Speech production is often studied using biomechanical models
of the vocal tract. Using a modal analysis procedure, it is shown
that the dynamical degrees of freedom for any such model can be
extracted by projecting the state trajectory at each point in
time onto a local eigenmode basis. The resulting time-varying
representation generalizes principal-component and linear-system
models of articulatory dynamics, and the dimensionality of the
system can be reduced appropriately by retaining only those
eigenmodes that make a significant contribution to the overall
movement. Simulation results are given to illustrate the
procedure.
|
