Static hand gesture recognition with 2 kinect sensors

In this paper, we propose a robust static hand gesture recognition algorithm that makes use of two Kinect sensors. This will be used to control an avatar in a decision aid for rheumatoid arthritis patients who have a difficult time using a standard keyboard and mouse interface. The sensors are placed on the left and right sides of a target sensing area, easily set up in a doctor's office or waiting room. The Kinects provide a rich point cloud, out of which gestures from a known vocabulary are recognized in real time. We use 6 point cloud descriptors simultaneously and employ the majority rule voting scheme to pick a "winner" gesture in real time. We achieve rotation invariance by using part of the forearm as a good indicator of hand orientation and aligning the hand with the world coordinate system origin. We evaluate the performance of the recognition system under various motion and rotation conditions.