Fingerprint matching based on weighting method and SVM

Fingerprint verification is an important biometric technology. In this paper, an improved fingerprint matching approach that uses weighting method and support vector machine (SVM) is presented. The traditional minutiae-based matching task is transformed to a classification task in the proposed approach using SVM. Furthermore, a new weight feature is introduced based on the distance between minutiae to supplement the minutiae information, especially for fingerprint images of poor quality. To give an objective assessment of the approach, both international and domestic fingerprint verification competition databases have been used for the evaluation. Experimental results show substantial improvements in the accuracy and performance of fingerprint verification.