problem with my GUI (I am currently doing hand signal recognition)

I am currently doing hand signal recognition, my problem with my GUI if he do snapshot he cant recognize well... but if i do load push button of the image it can recognize well.. I guess i need help for the snapshot that he can recognize well.. if you had any time please message me on skype myjonas_13 and lets talk about it there.. I am willing also to give rewards if someone solve my problem =(

[Walter Roberson]

I suggest you post a sample snapshot and a sample of an image that is recognized.

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

Answer 1

[Image Analyst]

I can guarantee any usable gesture recognition system won't be done in just a dozen lines. I can't help you with algorithm development - it would take way too much time - like weeks or months - not just a few minutes which is what we can typically spend on an "Answers" answer. So why don't you see how it's done by people who have already done this? You're not breaking new ground or doing cutting edge research here. It may or may not surprise you but many many people have done this before and you can see what all of them have done if you go here and read their papers: http://iris.usc.edu/Vision-Notes/bibliography/contentspeople.html#Face%20Recognition,%20Detection,%20Tracking,%20Gesture%20Recognition,%20Fingerprints,%20Biometrics Check out section 21.4. Pick your favorite algorithm and if you have programs related to MATLAB programming (not related to explaining how an algorithm works in some article, which we don't do) then you can come back here and we can help with your MATLAB syntax or other errors that you can't solve by yourself using normal debugger techniques. I hope the link helps you develop a successful gesture recognition system. Good luck.

[john john]

wow thanks with this.. naks hope i can learn on this asap =(

Answer 2

[john john]

can i contact you trough skype sir? so that we can discuss it much further?

i build a fix place set up like fixed background, amount of light, those images being recognize are those images i trained on SOM.. and those snapshot is from real time video capturing or snapshot being process =( i guess i need a code which reconsider or add variation of position of hand images to the network being train?

[Walter Roberson]

I cannot be contacted through skype or google voice.

You could drop by my office; you can't miss it, I'm right in the center of Canada.

[john john]

sad to say I am from Philippines =(

[Walter Roberson]

The only time you can see Major Major in his office is when he's out.

Answer 3

[john john]

but I notice those images I trained using SOM are different on snapshot.. I only train 10 variation of letter "a" ... how do I add data to my network that I trained? which for sure he for sure he can consider that its part or same letter pattern..

[Walter Roberson]

We can't do much to help you until you post sample images from both cases, as I asked 5 minutes after you posted your question.

[john john]

okay wait up..I am going 2 post it now..

[john john]

any reply sir Walter ROberson

[Walter Roberson]

I stayed late at my office, no-one showed up.

Answer 4

[john john]

http://i1065.photobucket.com/albums/u390/myjonas1990/fromsnapshot2.jpg

http://s1065.photobucket.com/albums/u390/myjonas1990/

cclick the link and you will see those photos...see the title tags for description of which use by train images and new snapshot..

the images is resize, grayscale and blackandwhite using this code..

wa = getsnapshot(vid);
x = wa;
y = double(x) / 1000;
data3 = imresize(y,[40 30]); 
data4 = rgb2gray(data3);
      
    %new
    img2=im2bw(data4,graythresh(data4));
    img2=~img2;
    BW2 = imfill(img2,'holes'); 
    z = BW2;
    z = z(:);
%load network, simulate
load trained_som_20x20m_10ke_30x40p_ae;
input = sim(network,z);
save('input_recognize.mat', 'input');

but it seems if any movement of hand which is not similar to the images i trained it wont detect.. =(