How To: Build A Successful Kinect Application

My primary function here at Vectorform has been researching development strategies for Microsoft’s Kinect motion sensor. This includes discovering and documenting best practices for use in our application design process. This is has been a necessity for us, seeing as how real standards have not been set by Microsoft at this time. Since there are independent developers out there experimenting with Kinect as a control platform, I thought that sharing some of these conclusions we’ve come too might help shorten the development time on some of your crazy hacks you’ve been brewing.

1. Bigger is better:
With Kinect, bigger is definitely better, and I’m not just referring to button size. We’ve been finding that finer movements tend to tire out users quicker, compared to larger, broader strokes. This doesn’t mean you need to space buttons out to different sides of the screen, though. We set up our applications so that the cursor will register just off screen when the arm is fully extended. The same process also applies to gestures, larger, slower movements are not only easier to track, but much quicker for users to learn and less taxing on the shoulder muscles.

2. Consider Space:
For optimal performance in full body tracking, the Kinect sensor needs an open space about 13 feet deep by 6 feet wide. You should always take into consideration the space available for your deployment.  In tight situations, full body tracking may be unusable and an upper body tracking approach may be more suited, since the users can be closer to the camera in these scenarios. In a very close camera scenario, individual hand tracking might be your only solution.

3. User Interaction:
In an application scenario, using complex multi-part gestures might be a considerable hindrance to user interaction. Multi-part gestures, while admittedly cool, are tough for your average user to get a handle on. A better solution is to build up on a pool of smaller, simpler gestures and combine those into simple two to three stroke gestures to create a set that your applications uses.

4. User Feedback.
Every intractable object within your application should have some sort of feedback, this will allow the user to easily understand and react to what is happening on screen. Combinations of both visual and audio feedback create a nice tactile feel that users can easily understand.

With yesterdays release of Kinect for Windows SDK Beta, we thought these quick tips might help. At Vectorform, we’ve put a good amount of effort into finding out these points, and while we don’t always stick with them, our biggest successes have followed all of them.