Improving Biometric Identification Performance Using PCANet Deep Learning and Multispectral Palmprint

Biometric technology is an emerging field of information technology that recognizes a person based on a feature vector derived from specific physiological or behavioral characteristic that the person possesses. In the last few years, several works in the field of biometrics got to improve the identification system performance rather than the traditional methods. So far, with the pace of rapid evolution in these works, new biometric modality (palmprint) is appearing to make the process of identification more efficient. There are a number of studies addressing the palmprint modality and the majority of these studies are mainly based on image captured under visible light. However, multispectral imaging technology has been recently used to improve the performance of biometric system. Furthermore, the feature extraction phase plays an important role in the biometric system. For that, several researchers are focused on methods used to extract the majorities of the characteristics that can discriminate each modality, which can decrease the intra-class variability and increase the inter-class variability. In this context and with the growing interest in biometrics applications, the studies in this chapter try to combine the multispectral imaging of palmprint and a new feature extraction method, called PCANet deep learning, in order to improve the system accuracy. To evaluate the performance of the proposed scheme, a database containing palmprint images was required. Thus, experiments were performed using two popularly databases: PolyU and CASIA databases.