Accuracy Comparison: Optoelectronic Motion Capture and Markerless System

The raw data acquired from motion capture system were processed with Smart Analyzer software (BTS Bioengineering, Milano, Italy). First, the 3D data were filtered and interpolated in case of missing data for short time. Then spatial-temporal parameters (cycle duration, cadence, gait speed, stance phase, swing phase, double-support phase, stride length and step width) and conventional kinematic parameters of traditional Davis marker-set protocols were computed.

The two videos were elaborated using OpenPose that returns a set of 25 2D keypoints coordinates for body pose estimation for each video. Key-points were located in relevant body landmarks and it were used to determine the 2D Cartesian coordinates on the sagittal plane and on the frontal plane. The data calculated with routines were filtered and interpolated in case of missing data. With respect to kinematic parameters, the segment and joint angles were measured from the estimated feature points of each joint. Spatiotemporal gait parameters were calculated using successive heel strike and toe-off events.