Clever Geek Handbook

Optical flow against the Haar cascade

intend to program the navigation of iPad apps with head movements (initially manual movements, but hands seem to be difficult to detect at the moment): left, right, up and down. So I am planning to use openCV and either detect the optical flow of the head area, or detect the user's ears and head with haar cascades (openCV comes with a fairly accurate xmls head and ear). Can anyone suggest any advice for using any solution? Will one of the mentioned need more power? Could it be easier to program ear and head recognition? I would like to avoid a lot of effort in the wrong direction and not have much experience in my current environment ...

thank you for any help, advice, ideas!

+3


source to share


3 answers


I would suggest using Haar cascades because optical flow is more expensive in computational time!



0


source


Is using native face detection in iOS 5 not an option for what you want? I think outside the box here, but I've been with OpenCV before and it still hurts ..

http://maniacdev.com/2011/11/tutorial-easy-face-detection-with-core-image-in-ios-5/



* The idea, of course, is that you apply this to directly input

0


source


One way would be to use hidden Markov models. There are many research papers (working prototypes) on how to use HMM for gesture recognition.

I find that ear classifiers are not as effective with a cluttered background. Please let us know if you have the opportunity to work!

0


source






More articles:


All Articles