Evaluation of Microsoft`s Kinect® for lower Limb Kinematics

Background: Human motion capture and analysis is widely used within a variety of purposes including assessment of surgical outcomes in rehabilitation stages. Currently, most clinical motion analysis centers use a marker-based system (MBS), these systems are costly and not available in many clinical settings. Recent studies evaluated Microsoft Kinect® sensor as a low cost and highly available motion capture system, but none of them compared it simultaneously to MBS while capturing subjects` several joints, on several planes, while walking.

Questions/Purposes: The purpose of this study was to evaluate Microsoft Kinect® sensor as a Marker Less System for motion capture and analysis of lower limb motion, and compare it to a state-of-the-art MBS. The objective was to investigate the system’s accuracy to simultaneously capture movement of several lower limb joints on several planes, while subjects walked freely.

Methods: Software for analyzing Kinect data stream was developed using Microsoft’s Visual Studio and Microsoft’s Kinect SDK. Subjects were asked to walk while gait data was simultaneously recorded by MBS system and Microsoft Kinect® sensor. Visual 3D (C-motion) software was used to calculate 3D joint angles of the MBS. Joint angles calculated by the two systems were compared using Root Mean Square Error (RMSE) calculation.

Results: The gait pattern recorded by the MBS and Kinect Kinect® sensor were similar. The calculated RMSE average was less than 4⁰.

Conclusions: Quantitative measurements of joint angles of the knee and hip can be acquired using one Microsoft Kinect® sensor with relatively good accuracy. The mean RMSE was less than 4⁰ which is clinically insignificant in most orthopedic interventions. The system can be advantageous for clinical use, at the pre- and post-treatment stages of rehabilitation. A future evaluation of the Kinect sensor should be performed among larger study populations.