Human hand identification with 3D hand pose variations

Several recent research efforts in the hand biometrics have focused on developing contact-free personal identification system. However, the acquisition of images from the unconstrained hands can introduce significant hand pose variations which severely limits the performance and applicability of these approaches. This paper presents a new approach to achieve significantly improved performance even in the presence of large hand pose variations. The proposed approach firstly estimates the orientation of the hands in 3D space and then attempts to normalize the pose of the simultaneously acquired 3D and 2D hand images. A new feature representation, namely SurfaceCode, is proposed for matching a pair of 3D palms. Multimodal (2D as well as 3D) palmprint and hand geometry features, which are simultaneously extracted from the texture details of the normalized 3D hands, are used for the matching. Individual matching scores are then consolidated using the weighted sum rule. Our experiments on the database of 114 subjects, with significant pose variations, achieve consistent performance improvement, both for palmprint and hand geometry features considered in this work.