Popularity of PCA in Face recognition and suitability for palm print recognition

Principle Component Analysis

Is an orthogonal linear transformation, that transforms the data to a new system, where it is stacked according the variance, for instance highest variance in the first column and second highest variance in the second column, and so on. PCA is theoretically the optimum transform for a given data in least square terms [1] and fundamental of Face recognition. In pattern recognition it is primarily used to reduce the dimensions in a data set by retaining the characteristics that are prominent. It is essentially used, either directly or indirectly in Face recognition. However Eigenfaces is what encompasses and supersedes it.

Eigenface [2]

As Turk says in his paper, developing a computational model of face recognition is quite difficult, because faces are complex, multidimensional, and meaningful visual stimuli and Eigenfaces is what extracts relevant information in a face image, encodes it as efficiently as possible, and compares on face encoding with a database of models encoded similarly.

The suggested method is:

1. Acquire the training set of images and calculate the eigenfaces that define face space.
2. To get the match, project the new image onto each of the eigenfaces created before.
3. Determine if the image is a face at all, by checking whether difference from the training set is within a threshold.
4. If it is a face, enforce greater threshold to confirm that it belongs to the same face.

This can be given inflowchart [3]

I found a related algorithm implementation in Matlab [4] and modified that to suit our purpose and modified working code is attached here with samples.

We, before testing its suitability for palm print recognition, tried on face samples where it showed a 0-9000 difference on 5 images from the same face (out of 20 used for training set) and > 14000 for a different face (controlling set) which means for a threshold of 10000 it is likely to have 25% success rate in identifying the face and 100% success rate in rejecting the wrong one. However, when it was adapted for palm prints results were futile.

There were many difficulties in obtaining a training set data base. FERET is the most popular data set [5], and to paraphrase its introduction as given in the site “The FERET program ran from 1993 through 1997. Sponsored by the Department of Defence’s Counter drug Technology Development Program through the Defence Advanced Research Products Agency (DARPA), its primary mission was to develop automatic face recognition capabilities that could be employed to assist security, intelligence and law enforcement personnel in the performance of their duties.”. However, they release dataset upon only a handwritten request [6] which is impossible in our case, hence downloaded Yale university data set and tried the algorithm on Face recognition [7].

[1] From, wikipedia

[2] Face Recognition Using Eigenfaces by
Matthew A Turk and Alex P pentland, copy of the paper is attached.

[3] http://openbio.sourceforge.net/resources/eigenfaces/eigenfaces-html/facesOptions.html
: last accessed 28 Feb. 08

[4] http://www.face-rec.org
: last accessed 28 Feb. 08

[5] http://www.itl.nist.gov/iad/humanid/feret/feret_master.html

[6] http://www.itl.nist.gov/iad/humanid/colorferet/home.html

[7] http://cvc.yale.edu/projects/yalefaces/yalefaces.html