Hmmm, I couldn't quickly find a good one either.
The trick is that if we have some signal representing a "position" then the first derivative will be "movement". If we sense the gravity acting on the Wiimote as it's "position" (orientation, in this case) then any changes in that will represent the forces acting on the Wiimote. For one thing; it's clear the thing won't move unless some force is acting on it, right?
Ok, I'm not the greatest digital filter wizzard but if we have this signal and pull it through a SVF we know for sure that the DC component of the signal and other very low ones (representing the orientation of the object at rest) will end up on the LP output. Once the orientation starts changing, assuming the change is in the range of the filter's frequency, this will make the change in the value appear in some way on the other outputs.
Where it gets hazy for me too is how that results in the derivative and that's where I hope somebody like Perry will help us but that's the rough outline of how I understand it works. Admittedly a vague outline in dimly-lit smoke but filters are hard stuff (at least for me). Maybe I can re-find that printout I had.
Apologies to the real filter wizards around here that are no doubt gnashing their teeth in frustration at this crude explanation :¬).
Kas.