DeepPalm-A Unified Framework for Personal Human Authentication

Different biometric traits have been proved to carry unique biological information of an individual. Although their use is always problem specific but the good performance of palm print represents a recent trend in this field. In this paper, we have proposed a novel generalized segmentation Network (PSegNet), that automatically categorize the palm print image obtained from multiple sensory resources and then detect the fixed size ROIs accurately. To best of our knowledge, this is the first attempt, of an end-to-end trained object detector, inspired by Deep Learning technique namely faster R-CNN (Region based Convolutional Neural Network) has been employed to detect and localize the position of palm. Then, the ROI image is well enhanced and transformed into illumination invariant visual representation. Over this, hand crafted features are patch-wise computed and the sub-patches are recursively matched using deformable dense matching technique called as Deep-Matching. The experimental results are examined on three publicly available palm print databases namely Poly-U, CASIA, and GPDS-CL1 databases. The proposed approach achieves equal error rate (EER) of 0.083%, decidability index (DI) of 2.94 and rank-1 identification rates (CRR) of 100% that justifies the role of CNN based ROI segmentation and multi-scale Deep-Matching criterion.