Personal Verification System Using Hand Modalities

Hand shape recognition is a popular biometric technology that is widely emerging especially in medium security applications such as access control. In this paper, a personal verification system based on hand shape, palmprint and fingers modalities is proposed. A refined SIFT (Scale Invariant Feature Transform) matching algorithm is proposed on hand and palmprint images in order to refine false matched keypoints as much as possible. In fact, SIFT descriptors have proven their efficiency in pattern recognition area and particularly in object recognition field. The main contributions of this work are as follows. Concerning the first one, it pertains to the localization of keypoints on the contour of the hand in order to describe the shape of the hand. As for the second one, it consists in two levels of SIFT matching refinement of hand shape process and one level for palmprint process, in order to refine as much as possible the number of matched keypoints and eliminate false matches. Regarding the third one, it involves the fusion scheme of the proposed verification system including the hand shape and the palmprint modalities at matching score level with fusion of fingers surfaces at decision level. Our experiments were assessed on 1150 hand images of IITD hand database, acquired from 230 subjects. The obtained results are promising and competitive to other biometric systems based on hand images.