Robust movement segmentation by combining multiple sources of information

One of the first steps in analyzing kinematic data is determining the beginning and end of movement segments. This is often done automatically on the basis of one parameter (such as a speed minimum) and subsequently corrections are made if visual inspection of other kinematic parameters suggests that the obtained value was incorrect. We argue that in many cases it is impossible to find a satisfactory endpoint for all possible movement segments within an experiment using a single parameter because the intuition about the end of a segment is based on multiple criteria. Therefore by taking the maximum of an objective function based on multiple sources of information one can find the best possible time point to call the endpoint. We will demonstrate that this Multiple Sources of Information method (MSI-method) for finding endpoints performs better than conventional methods and that it is robust against arbitrary choices made by the researcher. Using it reduces the chance of introducing biases and eliminates the need for subjective corrections. Although we will take goal directed upper limb motion as an example throughout this paper, it should be stressed that the method could be applied to a wide variety of movements.