查看文章

Palmprint recognition has become popular and significant in many fields because of its high efficiency and accuracy in personal identification. In this paper, we present a scheme for palmprint features extraction based on deep convolutional neural network (CNN). The CNN, which naturally integrates low/mid/high-level feature, performs excellently in processing images, video and speech. We extract the palmprint features using the CNN-F architecture, and exactly evaluate the convolutional features from different layers in the network for both identification and verification tasks. The experimental results on public PolyU palmprint database illuminate that palmprint features from the CNN-F respectively achieve the optimal identification rate of 100% and verification accuracy of EER = 0.25%, which demonstrate the effectiveness and reliability of the proposed palmprint CNN features.