From: Sahil on
I'm working on a dynamic gesture recognition algorithm. I'm trying to detect zooming in/out gestures (like the pinch gestures you do on an iPhone). I'm at a stage where the edge of the hand is perfectly detected. The detected hand in one image frame looks like this.

http://img715.imageshack.us/img715/867/56205285.jpg

My next task is to detect the fingertip only, since that is the part of the hand that is actually moving. Can someone point me in the right direction towards solving this problem? I assume that using the generalized Hough transform for detecting a curve which looks like my fingertip might help. Would it?

Thanks a lot.
From: Rune Allnor on
On 20 Mar, 08:30, "Sahil " <sahilsapre-remove-t...(a)ymail.com> wrote:
> I'm working on a dynamic gesture recognition algorithm. I'm trying to detect zooming in/out gestures (like the pinch gestures you do on an iPhone). I'm at a stage where the edge of the hand is perfectly detected. The detected hand in one image frame looks like this.
>
> http://img715.imageshack.us/img715/867/56205285.jpg
>
> My next task is to detect the fingertip only, since that is the part of the hand that is actually moving. Can someone point me in the right direction towards solving this problem?

Don't know if it is the *right* direction, but one approach could be
to check the outline for curvature: The longest segment of the
outline
with the highest accumulated curvature is one obvious candidate for
the finger tip. If you also is able to narrow down the curves as
convex or concave (i.e. you can estimate to which side of the curve
the interior of the hand itself is), you ought to get a long way.

Rune