Accuracy Evaluation of Modern Methods of Motion Capture

Purpose: The assessment of limb movement trajectories and joint ranges of motion (ROM) are critical components in diagnosing and treating movement deficits and injuries. Recent advances in the field of automated image analysis permit the identification of joint center locations using markerless anatomical detection. We evaluated the accuracy of a virtual reality system and synchronized web cameras against a research-grade system. Methods: Subject volunteers (n=31, m=14 f=17, age=21.2±3.2) completed a series of three scripted movement sequences, which were displayed to the subject using a custom virtual reality platform. The scripted sequences were designed to replicate the joint ROM and balance maneuvers typically used during a physical therapy session. While the subjects mimicked the movement sequence, we measured the joint xyz locations using, the VR headset, a series of webcams, and a research grade 10-camera system. The systems were temporally synchronized using a plunger to begin the recording period in the VR system, which illuminated an LED in camera field of view and sent a TTL pulse to the research system. The 2D pose estimates from the webcam videos were triangulated to 3D using a modified direct linear transform. For each subject and trial, we determined the difference between the joint center locations obtained from the webcam and VR systems vs the research system. Significance: The accurate assessment of real-time movement kinematics may permit the incorporation of lower body measures (currently non-instrumented) into immersive VR platforms that currently assess ROM and muscle forces for the arms and torso.