Direct fit of a theoretical model of phase transition in oscillatory finger motions

This paper presents a general method to fit the Schöner-Haken-Kelso (SHK) model of human movement phase transitions directly to time series data. A robust variant of the extended Kalman filter technique is applied to the data of a single subject. The options of covariance resetting and iteration within recursion were used to obtain time-dependent estimates of both the alpha and beta parameters in the SHK model. Comparison between transition onset time and the time at which /beta(t/T)/alpha(t/T)/ becomes critical indicates that the transitions are advanced by noise. The method can be extended to handle non-normal data and generalization across subjects and/or experimental conditions.