Invariant Hand Biometrics Feature Extraction

Hand biometrics relies strongly on a proper hand segmentation and a feature extraction method to obtain accurate results in individual identification. Former operations must be carried out involving as less user collaboration as possible, in order to avoid intrusive or invasive actions on individuals. This document presents an approach for hand segmentation and feature extraction on scenarios where users can place the hand on a flat surface freely, without no constraint on hand openness, rotation and pressure. The performance of the algorithm highlights the fact that in less than 4 seconds, the method can detect properly finger tips and valleys with a global accuracy of 97% on a database of 300 users, achieving the second position in the International Hand Geometric Competition HGC 2011.