My main goal is to keep it as simple as possible. I would like to keep this down to a webcam and at best two torches. But I realize that it doesn't work with the distance, so I have to make it a bit more complex, but I'd rather not touch IR.
Other wise the Wii-Remote solution would solve all my problems.

But I just realized that when the observer moves left and right in front of the screen, he always turns towards the screen and the webcam, if he's directly in front of it, he wont turn he's head away. (or shouldn't)
So maybe I can just account for the amount of head turning when he's moving to the side, that the distance stays steady enough.

It's not perfect, because the angle changes when getting closer to the webcam, but then again, this is not that precise, and it doesn't have to be, just close enough.