Favorite technique for hand gesture recognition

Question

There is a lot of research going on about gesture recognition. I figured I would narrow this down to the topic of hand gesture recognition (i.e. stationary hand positions, up to as complex and dynamic as sign language recognition).

Considering the image processing techniques available in real-time, such as blob detection, edge detection, point of interest tracking, etc. Coupled with Hidden Markov Models and other comparison AI, what techniques/algorithms would you use to do real-time motion tracking and gesture recognition?

Answer 1

i think for being accurate you have to combine all these techniques..

i did such a think with wii-mote, it recognizes the hand move but not gesture..

we used hidden markov model and it was succesful in real time...

maybe it is not a real answer but you cant know without trying..

Answer 2

well i think you can use active shape models or active appearence models i remember seeing some papers about it

Answer 3

I remember previous colleagues of mine have worked on a similar problem. They have published a conference paper on this topic: Framework for a portable gesture interface. Hope this helps.