Testing the Microsoft kinect skeletal tracking accuracy under varying external factors

Abstract

Focusing on its possible use in motion analysis, the accuracy of the Microsoft Kinect was investigated under various external factors including relative position, external IR light, computational power and large nearby surfaces. Two different experiments were performed that either focused on a general situation in an open room or when seated at a table. Results indicated that a large number of factors significantly affect the measurement error, but with only minor effect sizes, where the relative position and orientation have shown to be most influential. Additionally, body movement and increased depth contrast (i.e. isolation from surrounding objects) are believed to increase the accuracy of the skeletal tracking process.