Learning Performance Models For Post-Stroke Therapy From Demonstration
The use of robots and games is becoming a popular trend in stroke rehabilitation research. Despite the acknowledged importance, however, research on evaluating the patientís qualitative motor performance, customizing the initial task workspace based on the patientís performance, and generating visualized performance reports of autonomous sessions for the therapist have received little attention. The essence of these is to model the therapistís decision criteria to evaluate the qualitative motor performance and to use such evaluation to assess difficulties in the different regions of the task workspace for the individual stroke patient. In this talk, first I will motivate the needs of such models by introducing the results of a single-subject case study, and then propose a computational framework to learn the necessary models from therapist's demonstration. The experiment results with two patients and two therapists suggest that the proposed framework is promising.
Hee-Tae Jung is a PhD candidate at the College of Information and Computer Sciences, University of Massachusetts Amherst, and a member of the Perceptual Robotics Lab. He received his BS in Computer Science from Yonsei University, Korea in 2007 and MS in Computer Science from Stanford University, USA in 2009. His research focus is on developing and analyzing intelligent assistive technologies for the population with special needs. He is the recipient of the Robin Popplestone Scholarship and the Graduate School Dissertation Research Grant. He was selected as an ACM/IEEE Human-Robot Interaction (HRI) Pioneer in 2015 and was an elected PC Co-Chair for the ACM/IEEE HRI Pioneers Workshop 2016. He served as an editorial assistant for the Journal of Robotics and Autonomous Systems from 2013 through 2014.