An efficient fingerprint matching using continuous minutiae template learning

Abstract

Biometric samples acquired by any sensor are highly dependent on the behavior of the user with that scanner. No two fingerprint samples acquired at different instant of time are to be exactly same. This could be due to the sensor properties, translation, and rotation along with smudging effects. However, some of these effects can be minimized by using multiple images of the same trait and thereby improving the template or feature vector by assigning higher weights to the prominent features. This paper deals with the problem of improving the quality of the template of fingerprint with the help of multiple fingerprint samples acquired from the same finger of the subject at different instant of time. It has used the method of continuous minutiae template learning. The proposed method has been tested on FVC-2004 DB2 A database and has shown significant increase in the accuracy of the fingerprint-based biometric system if one adopts the proposed template learning method.