A Kinect-Based Motion Capture Method for Assessment of Lower Extremity Exoskeleton

摘要

Rehabilitation exoskeleton provides a new method for therapy of stroke patients, but it needs a portable and effective method to analyze the validity of exoskeleton rehabilitation training. This paper proposes a motion capture system based on Kinect to measure the variation of joint angles of a patient's lower extremity during rehabilitation training. Comparing the measured angles with the input rehabilitation trajectory, human-machine coupling property of exoskeleton can be achieved to analyze its validity. The system used image sequence motion detection algorithm based on markers and arranged eight marker bars on the surface of lower extremity. Then it got bars' spatial direction vectors by clustering algorithm (DBSCAN) and least square method. With lower extremity 5-bar model built, the system finally gained the joint angles. Meanwhile an experiment was designed to verify the precision of Kinect motion capture system. Results showed that the maximal static error was 2.76° and correlation coefficient of dynamic track was 0.9917. This proved that the Kinect motion capture system was feasible and reliable to provide parameter foundation to assess the validity of exoskeleton.