Currently the 'best' solution seems to use Wiimotes with IR-LEDs on the camera.
Check out these projects:
http://molviz.cs.toronto.edu/molviz/ (implementation in pymol)
http://code.google.com/p/immersive-viz/ (related)
http://code.google.com/p/htdp/ (related)
http://blog.insightvr.com/?p=16%23more-16
For WiiMote details, see another feature request.