Well,
I've got 2 pennies to share:
I was thinking getting consistent blobs will be a pretty challenging
task in itself. If it's for an installation, you can setup your lighting and
background to make your life easier, but if you want to use this for website, where
everyone has different cams/lights/backgrounds, it is going to get tricky.
After that, if you can keep track of blob ids, then use a blob as a 'cursor', then
you could easily implement mouse gestures that are controlled by a blob.
Didier Brun wrote a nice mouse gestures as3 library some time ago.
Actually, why bother tracking a constant blob. I think using OpenCV's Optical Flow
will give you the overall direction of movement. You might use your hand, your head, whatever. It might be easier to use play with openCV in a different programming environment, like openframeworks or max/msp/jitter using the cv.jit library. Whatever helps you prototype quickly, then get back to as3's openCV implementation.
2.
There might be better ways than this, but the paper looks pretty interesting and maybe
there are some ideas to use in your approach:
The Representation and Recognition of Action Using Temporal Templates
HTH,
George