[sldev] Body motion and facial expression tracking, Microsoft did it

[Carlo Wood]

On Sat, Jun 06, 2009 at 03:33:54AM +1000, Tateru Nino wrote:
> Maybe it isn't really about handshakes, and more about general
> lining-up-of-body-parts between avatars? :)

I'd implement this by tracking the *position* of the main
points (hands, head top, mouth, shoulder, crotch, etc),
and then define the skeleton to be the distances between
those points.

Thus, if the person recording a movement brings his/her
hand to her mouth - that particular distance would go
to zero.

When playing back, I'd then morph the skeleton such that
shorter distances get higher priority. Thus, a distance of
zero becomes garanteed zero, while a distance of 1 meter
may become 0.9 or 1.1 meter if needed. I'm sure this
can be achieved with rather simple matrix operations.

In couple animations you can simply extent this model
with the main points of the second person. That way
you are garanteed that a handshake results in a handshake,
a kiss in a kiss and a fuck in a fuck.

-- 
Carlo Wood <carlo at alinoe.com>