Analysis of Inverse Kinematics Solutions for Full-Body Reconstruction in Virtual Reality

Many Virtual Reality (VR) applications usually visualize only the VR controllers or floating hands. However, to create an immersive experience, a full-body avatar is essential. We reconstruct a full-body avatar by tracking the position and orientation of the head, hands, feet, and hip. To track the arm movements, the user has to hold two HTC Vive controllers. Additionally, the user has to bind Vive trackers to both ankles and the hip. We apply some of the most popular Inverse Kinematics (IK) methods to estimate the full-body pose. We perform parameter optimization to analyze the damping constant, the maximum number of iterations, and error value for the position as well as rotation. We made several tests between the IK methods in terms of the accuracy and the time to solve the IK problem. The results show that Damped Least Squares (DLS) outperforms the other methods. We furthermore conducted a user study to evaluate the subjective quality of the DLS method. Evaluation results show that the motion reconstruction for lower-body is very accurate; however, for the upper-body, some inaccuracies can occur. Such a motion reconstruction approach can be used in VR exergames, e.g., users can learn different poses while observing the movements in a virtual mirror and by looking down towards their own body.