In a simple motion capture system for game input you might only need to capture 4 points. These would be: Head position (if this is nessesary), Left hand position, Right Hand position and Torso position. With these you can use them as input for camera rotation (head), motion (torso) and weapon position (left and right hands).

This still might be too complex and expensive for a home system (though if you had colour coded gloves and markers to aid in it, it might be easier).

I did some work with an artist a few years ago that made a virtual 3D maze, using just a few digital video cameras and a computer. Each camera could be assigned several polygons that if it registered a difference in a prerecored background image within the polygon it would signal the system that a user might be intersecting a virtual wall. You could link polygons on different cameras to produce a 3D area that contituted a wall. If all linked polygons registered that a person was inside the linked shapes then it sounded a tone to indicate that the person had entered a wall (it was quite amusing to wach people move around the maze waving their hands around to find the "walls"). Part of it was also for sientific purposes and we hade to collect some biometric infomation from the person (in the form of breathing and heart rates).

I have though of recreating this system with off the shelf web cameras, however I can't do the programming for it.

This system could be used to make a new form of controler for a video game as you could have virtual "buttons" where the user must move their hand into a certain area in space to use the button (like touching somthing infront of them). You could also track a coloured object in 3D (using the cameras to calculate it stereoscopicaly). This could be used as an input into a game too.

And for feedback for enemies hitting you, etc., you could wear a suit with taser-type electrodes at various points which zap you if you aren't fast enough.