Saurabh Aggarwal

Dr. Kamlesh Tiwari

The human fingerprint is one of the most convenient and mature biometric traits. It’s potential to uniquely identifying humans was noticed first in 1858 by Sir William Herschel. The first research paper about automatic recognition of fingerprint images, by Trauring, came in 1963.

Fingerprints are a well-accepted biometric trait for personal authentication because of their reliability, individuality and persistence. Fingerprint patterns differ from person to person, even in monozygotic twins. Fingerprinting for identification enjoys broad acceptance by the public and law enforcement communities. The primary disadvantages of the method include its high user cooperation requirement, its susceptibility to fingerprint spoofing, and potential deterioration in the quality of the acquired fingerprint sample.

Various fingerprint-based recognition systems have been proposed in literature. They all can broadly be categorized into texture based and minutiae-based methods. Despite fingerprints being one of the most popular and widely accepted biometric traits, the performance of fingerprint-based recognition systems is not yet saturated, especially in large scale deployments.

As fingerprints are biological traits, they are susceptible to aging. Fingerprints that are acquired at the time of registration become spatially distant over time from the ones taken from the user for authentication. This may lead to failure in determining that the two fingerprints are from the same user. There is thus a need to learn from successful biometric authentication instances, to improve fingerprint templates in hope of better recognition accuracy.

The paper proposes a minutiae confidence based fingerprint template learning approach which utilizes Machine Learning and Image Processing Techniques to enable Priority based Minutiae Matching. This approach helps compensate for the effects of aging on the fingerprint template.