CLASSIFICATION OF NONLINEAR DYNAMICS OF HUMAN POSTURE USING SUPPORT VECTOR MACHINES

Human postural control can be modeled as a complex, nonlinear multi-sensory and multi actuation system. Postural control requires a balance between various sensory inputs and the corresponding response in a motor control framework. Various motor control disorders result in a propensity for fall. In this paper, a classification scheme using support vector machines (SVM) is investigated to classify individuals between healthy and those prone to balance disorders including falls. First, a computational model is described to illustrate the proposed approach. It is shown that using SVM, in conjunction with L-2 Norm associated with the phase space trajectories, it is possible to classify fallers and non-fallers. This metric can then be utilized for clinical application upon further evaluation. This may benefit clinical evaluation of Parkinson's subjects by complementing the widely used UPDRS scale.