A basic sketch to experiment with 3D tracking with Kinect and Processing.
Nothing complicated: once the user is calibrated the camera is focusing on him and switching focus point every 5 seconds. Focus points can be any joints of the 3D skeleton (head, torso, hips, hand, etc...). After each change of focus the camera then makes a random motion (dolly, pan, zoom, etc...). If the user is detected but not calibrated the camera focuses on the user's center of mass, if the user is neither detected nor calibrated then the camera focuses few meters away in front of the Kinect.
More info here: http://blog.goodthink.biz/kinect-tracking/