Contactless Palm Verification Using Siamese Neural Networks and Local Binary Pattern

Abstract

Biometric authentication is the confirmation of whether people are really the person they claim by using their physiological or behavioral characteristics. Palm verification is one of the most widely used methods in biometric verification. The COVID-19 (Coronavirus Disease 2019) pandemic emerging in the last months of 2019 has increased people's sensitivity to contact with objects of common use. In the study, Hong Kong Polytechnic University Contactless 3D/2D Dataset (Version 1.0) (PolyU Contactless Database 1.0) and Siamese Neural Networks (SNN) were used for validation. SNN trainings were carried out using a total of 34,692 pairs of images, of which 3,540 were "similar" and 31,152 were "dissimilar". Testing of the study was carried out using a total of 32,037 input samples, 885 of which were "real" and 31,152 were "fake". In the present study, the validation results were obtained using the palm images directly and the validation results were obtained by using Local Binary Pattern (LBP) as a pre-process. Then, these results were compared with each other. The results of the study show that the use of LBP as a pre-process significantly improves the validation success. In the study, while the Equal Error Rate (EER) obtained by using the palm images directly was 0.1277, the EER obtained by using the LBP as a pre-process was 0.0938.