I'm looking for the most efficient way to run in Kinect Sports and by that I mean the least fatiguing configuration of leg \ hand movement that corresponds to a fastest avatar movement in game.
I recently observed that running like hell does not necessarily imply that avatar's speed is a highest possible, but lifting knees higher helps achieve that state.
Any tips would be appreciated.
The short version is, yes, focus on raising your knees, and making exaggerated movements. The further away from your body and the more exaggerated the motion, the better. The following video shows a guy demonstrating some advanced techniques for each event in the Track and Field portion of Kinect Sports, along with his performance in the game and some commentary:
The longer answer follows :)
The Kinect works by trying to draw a 3-D model of your body based on a depth map generated by the Kinect's sensors. The Kinect has a color camera for taking images, and a monochromatic IR laser and IR sensor for depth perception. The following photo is from Wikipedia's Kinect article:
This is a processed photo of a person with a laptop on a table in front of them. The color coding indicates the distance from the camera as perceived by the Kinect. The "red" areas are closer to the Kinect sensors, while the yellow areas and green areas are further away.
As you can tell from this, the Kinect is likely to have difficulty distinguishing things that are very close together and at the same distance from the sensor. This person's right hand is clear once you've seen the original photo, but to the Kinect it could be part of the keyboard rather than a distinct object.
When you move your knees out in front of your body, you are giving the Kinect a better chance of "seeing" them as different objects at a different depth. Motions that move your hands or feet outside of the plane of your torso are also easier to detect. The Kinect is likely to have an easier time seeing your hands if you're holding your arms away from your chest, rather than keeping them next to your sides or up against your stomach, for instance.
It seems to me that it should be possible to "game the system" by changing the way you present yourself to the system. For instance, if you had something attached to your knee that ran upwards towards your waist (like a shin guard put on too high) the Kinect might see that and think it was your knee. You could move your leg up a relatively short distance and the Kinect might interpret that as you bringing your knee up rather high. Since you can make this motion with less effort and considerably faster, it seems like you'd have an advantage.
I also wonder if it would be easier for arm and leg motions if your arms and legs were high contrast versus the rest of your body. The Kinect's depth sensor is monochromatic, so slightly different colors wouldn't make any difference, but changes in the brightness of what you're wearing might. You could try wearing a shirt that is long sleeved and had a black chest with white sleeves, or pants that had a different color below the knee.
