University of Louisville
Motion assessment using Segment-based Online Dynamic Time Warping (SODTW) during social Human Robot Interaction (HRI) with Zeno robot
Grade Level at Time of Presentation
Senior
Major
Electrical and Computer Engineering
KY House District #
3
KY Senate District #
36
Faculty Advisor/ Mentor
Dan Popa, PhD; Sumit Kumar Das, PhD, Moath Alqatamin, PhD
Department
JB Speed School of Engineering - Electrical and Computer Engineering Dept.
Abstract
Robots are proving to be vital to assisting healthcare clinicians and patients in rehabilitation through imitation learning. By evaluating the motion imitation quality of a user as compared to a robot, the diagnosis of Autism Spectrum Disorder (ASD) can be inferred. The proposed Segment-based Online Dynamic Time Warping (SODTW) algorithm can be used for understanding repeated and cyclic human motions. By directing the user to imitate the motion of the robot, both human motion and robot motion can be analyzed through the SODTW algorithm. The collected data is analyzed to create more comfortable motions for users with ASD via tests conducted with neurotypical subjects. Recording cyclical data from both human markers and pre-programmed robot motion leaves room for a challenging metric to be determined on how to use it for diagnosis or treatment. Our proposed Reaction Sequence Index (RSI) investigates the effects of a human learning a task through hand motion and calculates their level of impairment. Through experimental joint angle analysis of plotted waveform data, we determine the sequence index in the measured time series where the human imitates robot motion the best. The motor performance of patients with disabilities can be monitored through this index, as studies have shown that subjects with ASD have longer reaction times than neurotypical subjects. The results from experiments of interaction by our social robot Zeno and human subjects prove that the algorithm can be used to design adaptive robotic therapies through cyclic motion teaching.
Motion assessment using Segment-based Online Dynamic Time Warping (SODTW) during social Human Robot Interaction (HRI) with Zeno robot
Robots are proving to be vital to assisting healthcare clinicians and patients in rehabilitation through imitation learning. By evaluating the motion imitation quality of a user as compared to a robot, the diagnosis of Autism Spectrum Disorder (ASD) can be inferred. The proposed Segment-based Online Dynamic Time Warping (SODTW) algorithm can be used for understanding repeated and cyclic human motions. By directing the user to imitate the motion of the robot, both human motion and robot motion can be analyzed through the SODTW algorithm. The collected data is analyzed to create more comfortable motions for users with ASD via tests conducted with neurotypical subjects. Recording cyclical data from both human markers and pre-programmed robot motion leaves room for a challenging metric to be determined on how to use it for diagnosis or treatment. Our proposed Reaction Sequence Index (RSI) investigates the effects of a human learning a task through hand motion and calculates their level of impairment. Through experimental joint angle analysis of plotted waveform data, we determine the sequence index in the measured time series where the human imitates robot motion the best. The motor performance of patients with disabilities can be monitored through this index, as studies have shown that subjects with ASD have longer reaction times than neurotypical subjects. The results from experiments of interaction by our social robot Zeno and human subjects prove that the algorithm can be used to design adaptive robotic therapies through cyclic motion teaching.