Regression modeling of motion with endpoint constraints

A statistical model is described for the prediction of reaching motions using motion capture data on a variety of individuals performing reaches to a range of targets. The modeling approach allows for various inputs such as the stature, age and the location of the target to be specified and then computes the predicted trajectories of the kinematic chains of body markers necessary to place an object exactly at the specified target. Functional regression methods for modeling time-varying angles and other quantities as well as trajectories are described. A new parameterization of posture is described that facilitates the satisfaction of specific endpoints such as placing an object at a target. The methodology is illustrated with an application to two-handed standing lifts. Copyright (C) 2003 John Wiley Sons, Ltd.