Yesterday I had the pleasure of speaking about the Kinect for the fifth time, this time at the North Dallas .NET User Group. NDDNUG is a great group. Even with it pouring outside, we had a decent turnout.

This was the first time since the 1.0 release I’ve presented on it. What I didn’t expect was how much shorter my presentation was due to how much work has gone into simplifying the .NET Kinect development experience. Bit twiddling on depth, MTA’s, DMOs, and much more are just gone from the experience. The team deserves some serious kudos for this.

The timing was auspicious. The Kinect 1.5 SDK includes many new features:

Face Tracking. This is huge for those trying to do animated films on the cheap. This will map a person’s eyebrows, mouth position and more onto a 3D mesh. You smile and your puppet self will smile.

Kinect Studio. Record and playback of Kinect data (No more jumping up and down just to write software.. though this was great exercise). This is going to make debugging key scenarios much easier.
Seated skeletal tracking. Seated skeletons will no longer dissolve into painful yoga poses.

Joint Orientation. The Kinect can tell you about rotated joints such as the rotation of the wrist. Good for playing twister and revving virtual motorcycles!
Improved green screen effect (mapping RGB to depth) most noticeable in the speed of tracking (from about 10ms to 2ms), though it still comes with a halo of noise.

Additional languages for speech recognition.

For new users and old, the quality of the Kinect developer samples and especially the Kinect Explorer (the key sample application to demonstrate Kinect features) have been greatly improved. Also released is a well written Human Interface Guidelines document. This 70 page PDF gives insight into good Kinect HCI for both audio and kinetic inputs. It also lends a domain language with which to discuss Kinect. What the guide doesn’t do is give technical details on how to develop gestures.

Developers are still hoping a gesture recognition and possibly a facial recognition SDK are in the works, but there isn’t a clear indication if these are in the works.

A couple people asked me about slides and code. Code is available on bitbucket and slides are here.