Reinforced Deep Learning for Verifying Finger Veins

Abstract

Recently, personal verifications become crucial demands for providing securities in personal accounts and financial activities. This paper suggests a new Deep Learning (DL) model called the Re-enforced Deep Learning (RDL). This approach provides another way of personal verification by using the Finger Veins (FVs). The RDL consists of multiple layers with a feedback. Two FV fingers are employed for each person, FV of the index finger for first personal verification and FV of the middle finger for re-enforced verification. The used database is from the Hong Kong Polytechnic University Finger Image (PolyUFI) database (Version 1.0). The result shows that the proposed RDL achieved a promising performance of 91.19%. Also, other DL approaches are exploited for comparisons in this study including state-of-the-art models.