Kinect v2 accuracy as a body segment measuring tool

摘要

Mild traumatic brain injury (mTBI) and/or concussion is the leading cause of injury related to recent U.S. military conflicts. Among other symptoms, injured service members often present with impaired balance. Several field-expedient test batteries have been developed to evaluate balance deficits in this context; however, such tests suffer from limitations associated with human observation and scoring. In order to address these limitations, we have developed the Automated Assessment of Postural Stability (AAPS) system, a computerized balance measurement tool which automates the frequently-administered Balance Error Scoring System [1], [2]. The AAPS is based on the Microsoft Kinect v2, a markerless, portable and low-cost alternative to laboratory-grade motion tracking systems. This hardware integrates an array of sensors including an HD camera, infrared, and depth sensors. The Microsoft proprietary skeletal tracking algorithm [3] estimates coordinates for up to 25 body joint centers which are used to recreate a stick model representation of the subject. All tracking data is encapsulated in a “body frame” stream generated at a variable frame rate of up to 30 fps.