Integration of Deep Hashing and Channel Coding for Biometric Security and Biometric Retrieval

In the last few years, the research growth in many research and commercial fields are due to the adoption of state of the art deep learning techniques. The same applies to even biometrics and biometric security. Additionally, there has been a rise in the development of deep learning techniques used for approximate nearest neighbor (ANN) search for retrieval on multi-modal datasets. These deep learning techniques knows as deep hashing (DH) integrate feature learning and hash coding into an end-to-end trainable framework. Motivated by these factors, this dissertation considers the integration of deep hashing and channel coding for biometric security and different biometric retrieval applications. The major focus of this dissertation is biometric security, wherein deep hashing is integrated with channel coding to develop a secure biometric authentication system. In this system, multiple biometric modalities of a single user are combined at the feature level using deep hashing (binarization). A hybrid secure architecture that combines cancelable biometrics with secure sketch techniques is integrated with the deep hashing framework, which makes it computationally prohibitive to forge a combination of multiple biometrics that passes the authentication. The integration of deep hashing and channel coding not only finds application in biometric security but it can also be extended to different biometric applications. To this end, the integration of deep cross-modal hashing and error correcting codes has been extended to improve the efficiency of attribute-guided face image retrieval. Additionally, the dissertation also presents a framework for cross-resolution (low-resolution to high-resolution) face recognition, and profile-to-frontal face recognition. A novel attribute- guided cross-resolution (low-resolution to high-resolution) face recognition system that lever- ages a coupled generative adversarial network (cpGAN) structure with adversarial training to find the hidden relationship between low-resolution and high-resolution images in a latent common embedding subspace is developed and presented. A similar framework that leverages cpGAN structure has been developed for a profile-to-frontal face recognition system. Finally, the performance of this cpGAN architecture for profile-to-frontal face recognition system has been evaluated and compared with a coupled convolutional neural network (cpCNN) and an adversarial discriminative domain adaptation (ADDA) network