I am working on sing language translator. I have 1500 images with 15 signs, 100 samples for each sign. The image size is 640 by 420.

As the total pixel number is huge, I was wondering how to reduce the number of features from those images to train matlab those images using PNN (probabilistic neural network). But before using PNN, I need to use PCA to reduce the size of the training matrix.

Can anyone help me to do that..?

"Nehal " <arnab620@yahoo.com> wrote in message <i4irug$66e$1@fred.mathworks.com>...
> I am working on sing language translator. I have 1500 images with 15 signs, 100 samples for each sign. The image size is 640 by 420.
> 
> As the total pixel number is huge, I was wondering how to reduce the number of features from those images to train matlab those images using PNN (probabilistic neural network). But before using PNN, I need to use PCA to reduce the size of the training matrix.
> 
> Can anyone help me to do that..?

I know the PRTOOLS toolbox has a PCA function: http://www.prtools.org/

Did you search on the file exchange? Lots of stuff there: http://www.mathworks.com/matlabcentral/fileexchange/

BTW: It is "principAL", not "principLE". Very different things, those two.