Abstract

The addition of a nonlinear restoring force to dynamical models of the speechgesture significantly improves the empirical accuracy of model predictions, butnonlinearity introduces challenges in selecting appropriate parameters andnumerical stability, especially when modelling variation in empirical data. Weaddress this issue by introducing simple numerical methods for parameterizationof nonlinear task dynamic models. We first illustrate the problem and thenoutline solutions in the form of power laws that scale nonlinear stiffnessterms. We apply the scaling laws to a cubic model and show how they facilitateinterpretable simulations of the nonlinear gestural dynamics underpinningspeech production.

 

Quick Read (beta)