Evaluation of Hand-Tracking Systems in Teleoperation and Virtual Dexterous Manipulation

The quality of robotic dexterous manipulation, in real or in virtual environments, relies on a fine control of the fingertips to perform stable grasps and inside-hand manipulation. In practice, teleoperating a robotic hand requires to capture the human hand configuration. If the user manipulates objects with fingertips, the acquisition of their motion must be accurate enough to produce realistic manipulation at the robot hand or its virtual avatar. In this context, one challenge is to accurately capture the motion of the human hand. The performances of three different hand-tracking devices are evaluated in this paper: two data gloves, the VRFree and the Manus VR, and a vision-based system, the Leap Motion Controller. To this end, the positions of the human hand joints and fingertips are captured while performing several tasks, with a high-precision motion capture system as reference, and with the tested devices. The accuracy of the measured joint angles and fingertips positions is compared for the different systems. Specific measurement configurations are considered by varying the hand orientation and the distances to the sensors. The strengths and weaknesses of these different systems are deduced from the experiments. This system review gives insights into the relevance of hand-tracking devices for remote robotic or virtual manipulation.