Nonminutiae Based Fingerprint Matching

Abstract

Fingerprint verification is one of the most reliable personal identification method. However, manual fingerprint verification is tedious, time consuming and expensive. Thus it is incapable to match the required speed and accuracy expected in modern world. Minutiae based and image based are two major approaches for fingerprint recognition. Image based approach offers much higher computation efficiency with minimum pre processing and proves effective when the image quality is too low to allow a reliable minutiae extraction. In this work, a novel method of image based fingerprint matching which uses the orientation field as a feature vector is proposed. The proposed method overcomes problems of shape distortion, variation in position, rotation and translation of image. Optimal core point detection of fingerprint (image) helps in improving accuracy of fingerprint matching system. Based on orientation feature, a fingerprint matching system is designed. The algorithm has been tested on two databases (10 subjects with five impressions from database available from neurotechnologija and 8 subjects with five impressions from FVC2002 D1_b). The performance of algorithm is measured in terms of receiver operating characteristics. For the neurotechnologija database at 0% false acceptance rate (FAR) the genuine acceptance GAR observed is 75% and at 21.1% FAR, GAR is 95%.For the FVC2002 database at 2.8% FAR the GAR observed is 80% and at 8.66% FAR the GAR is 97.5%. Computational Complexity of proposed algorithm is low and hence can be implemented as automatic fingerprint identification system.