MODELLING AND PERFORMANCE ASSESSMENT OF HUMAN REACHING MOVEMENTS FOR DISEASE CLASSIFICATION

Human arm motor control has been object of great investigation for several decades, during which some issues have been identified as themes of high interest. There is a wide number of studies on human motor control supporting the theory that reaching and pointing movements are the result of sequences of discrete motion units, called sub-movements. Evidence for the existence of discrete sub-movements underlying continuous human movement has motivated many attempts to "extract" them. Moreover, to analyze the strategy of the reaching movements, gained a great appeal in the rehabilitation field. In fact, understanding movement deficits following central nervous system lesions and the relationships between these deficits and functional ability, is fundamental to the development of successful rehabilitation therapies. The goal of sub-movement extraction is to infer the sub-movement composition of a movement from kinematic data. In the tangential velocity domain, a sub-movement is represented as a uni-modal, bell-shaped function. Determining the number, relative timing, and amplitude of sub-movements that most closely reproduce the original tangential velocity data is a non-linear optimization problem difficult to solve. The experimental observations suggest that sub-movements are ubiquitous but proof of their existence and detailed quantification of their form have been elusive. Although several sub-movement extraction algorithms have been proposed previously, all of them are subject to finding local, rather than global, minima and to producing spurious decomposition results. The first section of this thesis, propose a review on the decomposition methods developed until now and the several methodologies used to extract them. Furthermore, an hybrid sub-movement decomposition method is proposed, based on a robust expectation maximization (EM) constrained algorithm and a scale-space approach capable to overcome the limitations of the EM algorithm, which is a local maximum seeker. This representation allowed to explore whether the movements are built up of elementary kinematic units by decomposing each surface into a weighted combination of Gaussian functions. Finally, is proposed a new kinematic and electromyographic assessment of robot assisted upper arm reaching in hemiparetic subjects applying successfully the sub-movement decomposition method implemented to carefully analyze their motor and muscle strategy.