Another quick demo playing with open source drivers for the Kinect (openkinect.org)
A threshold is used on the depth-map to filter out everything but my hands, and then blob detection is used to locate their centers. This information is then used to scale and rotate an onscreen object.
Note that because the Kinect provides depth information, the object can be rotated on both its Z and Y axis. With a bit of work, a gesture could theoretically also be made to rotate along the X axis.
* sorry about the flickering, this is an artifact of screen recored I am using, and is not visible in actual use