Tracking Of Hand’s Posture And Gesture

The hand with its pose and gesture is not easy to track with cameras. It is mostly because of a high number of degrees of freedom and a complex shape. Our algorithm is based on three dimensional clouds of points issued from stereovision in order to keep the depth information. The complex and organic aspect of the hand is modeled by an articulated skeleton on which metaballs are attached. We reduce the difficulty of the problem by adding joint constraints to the model. The core of our program is an energy-based minimization that fits the model at each frame using the distance between the model and the clouds of points, the result at the previous frame is chosen as an initial guess.