Digital Improv Kinect Feb 2012 Update

We've improved how we measure and categorize full body gesture recognition with the Kinect for Digital Improv. After gathering sample user data, we realized that we needed a more forgiving and flexible solution. Now, rather than comparing total body position against sequences of posed keyframes, we're instead algorithmically determining which isolated, meaningful gestural units are being performed. Each signal tests only for the specific combination of angles and relative positions needed to make that claim, independent of what the rest of the body is doing.

This research project is led by Dr. Brian Magerko and has been contributed to by many other students both presently and over the past several years. All art shown here (i.e. both character and background) was made by Andrea Benavides.