Quantitative evaluation of parameters affecting the accuracy of Microsoft Kinect in GAIT analysis

Abstract

To date various commercial systems have been used in the GAIT analysis. These systems have some difficulties for clinical use, such as interfering with normal movement and high prices. The possibility of utilization of Kinect as a sensor for GAIT analysis has been studied in this research. The accuracy of Kinect in calculation of GAIT parameters such as lower limb joint angles, stride time, and stride length were computed during normal walking. The effects of the sensor's position and direction relative to the walkway were also investigated. The Kinect sensor was installed at different positions toward the motion path. In each position the data was recorded by both Kinect and a commercial motion capture system. After data recording and noise filtering, GAIT parameters were calculated. According to the results, the best position for the Kinect sensor is in front of the subject. In this arrangement, the error of stride time and stride length were 1.4% and 5.2% respectively. The root mean square errors of knee and hip joint angles in this position were 4.5° and 4.4°, respectively. The error of ankle joint angle was higher than the acceptable amount for clinical applications.