Riemannian Geometric Approaches for Measuring Movement Quality

Anirudh Som, Rushil Anirudh, Qiao Wang, Pavan Turaga; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2016, pp. 43-50

A growing set of applications in home-based interactive physical therapy require the ability to monitor, inform and assess the quality of everyday movements. Interactive therapy requires both real-time feedback of movement quality, as well as summative feedback of quality over a period of time. Obtaining labeled data from trained experts is the main limitation, since it is both expensive and time consuming. Motivated by recent studies in motor-control, we propose an unsupervised approach that measures movement quality of simple actions by considering the deviation of a trajectory from an ideal movement path in the configuration space. We use two different configuration spaces to demonstrate this idea - the product space S^1 x S^1 to model the interaction of two joint angles, and SE(3) x SE(3) to model the movement of two joints, for two different applications in movement quality estimation. We also describe potential applications of these ideas to assess quality in real-time.