Minimum Cost Fingerprint Matching on Fused Features Through Deep Learning Techniques

Biometrics is a scientific order that includes techniques for recognizing the individuals with their physical or behavioral attributes. The most widely recognized physical and behavioral traits of a man utilized for authentication are as per the following: fingerprint palm print, iris, retina, DNA, ear, signature, speech, keystroke elements, motion and hand-geometry. Among them, fingerprint is recognized and accepted as a universal trait. Fingerprint has a unique pattern that can easily recognize the individual. The key motivation behind this paper is to enhance the accuracy rate of human recognition by adhering to the factors like revocability, speed and with minimum cost. The high-level recognition rate is achieved through fusion of multi features like ridge endings and ridge bifurcations of fingerprint. Autoencoder (AE) is an unsupervised deep learning technique, exercised on the fused multi-feature representation template to address the various spoofing attacks and also achieve revocability. Minimum cost matcher (MCM) is applied to maximize the efficiency of the multi-representation system. The experimental results explain the high- level accuracy of the proposed system in human identification.