Project Team: Alex Vakanski (PI), Stephen Lee, Jake Ferguson

The current methods in the literature for representing human motions are based on modeling the movements at a single level of abstraction, either at a low-level (i.e., trajectory level) of abstraction or at the high-level (i.e., symbolic level) of abstraction. The proposed project will exploit the latest advances in recurrent neural networks for modeling human motions at multiple hierarchical levels of abstraction. The ultimate aim is to allow patients to perform rehabilitation exercises at home using a sensory system for capturing the motions, where an algorithm will retrieve the trajectories of patient’s exercises, will perform a data analysis by comparing the performed motions to a model of desired motions, and will send the analysis results to the patient’s physician with recommendations for improvement.