Palmprint and dorsal hand vein dualmodal biometrics

Nowadays, the fusion of different unimodal biometrics has attracted more and more attention of researchers who are dedicated to the real-world applications of biometrics. In this paper, we explored a dualmodal biometrics learning algorithm integrating palmprint and dorsal hand vein (DHV). Palmprint recognition has a considerable high accuracy and reliability, while the most significant advantage of DHV recognition is the so-called biopsy (liveness detection). To hybridize a dualmodal biometric algorithm combining the advantages of both methods, deep learning and graph matching were introduced to recognize palmprint and DHV, respectively. By adopting Deep Hashing Network (DHN), palmprint images can be encoded into 128-bit codes. Then, hamming distance was employed to represent the similarity of two palmprint images. Biometric Graph Matching (BGM) can obtain three discriminant features between two DHV samples. Feature-level fusion of DHN and BGM was conducted, and authentication was given by support vector machine. In this way, we can obtain the best experimental result with Equal Error Rate equal to 0, finally.