Wednesday, February 27, 2008

Neural Network based Pattern Recognition (Fingerprint)

This is a modified Self-Organizing Map designed specifically to learn fingerprints and can be used for fingerprint based verification and authentication. It can be used to recognize any kind of pattern.

What are the difference between this solution for fingerprint recognition and the others prevalent in the market? Firstly, the ones in the market are conventional. They try to get all the Minutiae points in the finger. Minutiae in a fingerprint are the various bifurcations and ridges that can be identified in a fingerprint. So in the whole of the fingerprint, only those minutiae are taken into consideration for authentication purposes. Which simply means you are not taking other information in the fingerprint that are very much unique to the fingerprint for authentication purposes. Some other solutions in the market take into account the directional vectors in the fingerprint for authentication purposes. Well this may be better than the solution that only looks for minutiae, but still we can do better.

A person when given a sample of fingerprint to verify against a stored fingerprint will not just look at the minutiae in the sample or the directional vectors in the sample but verify every detail in the sample against the stored fingerprint. Every point in the sample will be taken into consideration. Every dot in the sample would matter. The person will look at so much information that he or she would not need the whole sample to verify against the stored fingerprint, just a portion of the sample will do for the person to successfully verify against the stored fingerprint. That is what has been achieved through this solution.

Please watch the video below for further information. Happy watching!!!!



To see the complete video in one shot please go to http://video.yahoo.com/watch/2105236/

Thanks,
Vignesh Arjunan