Sensorimotor Synchronization - Modeling human response to step changes: Using continuous transfer functions

Abstract

The purpose of this thesis is to explore the possibility of continuous modeling of rhythm production. To do this it is assumed that the irregular time intervals of rhythm production can be regarded as irregular samples of a continuous process. Based on several preliminary experiments, system identification theory and previous research an experiment with step changes and ramps is conducted. Making use of averaging the noise observed in the preliminary experiments is reduced. A persistent overshoot is observed for the step changes and it is decided to evaluate only four transfer function models. The second order underdamped transfer function model with a zero and a time delay shows the best performance. Using different sets of data for validation and for estimation reveals only partial success of the model. The model ultimately fails the validation test and is thus discarded for the purpose of this thesis. Within the work with this thesis it is also found that clapping shows lower variability than finger tapping, which is explained by a slower error correction process. The overshoot observed in the step changes is attributed to the necessity of keeping phase relation and not because of an initial over estimation of the new rhythm.