Identity recognition is becoming increasingly important in many applications including access control and e-commerce. Current approaches for identity recognition are based on passcards or PIN numbers that can be stolen or forgotten. The use of biometrics (e.g., face, fingerprints, voice patterns, palmprints, iris images, etc.) will improve security, since biometrics are integral to a person. Recognition includes verification (authenticating or rejecting a claimed identity) and identification (matching a presented biometric to one of several in a database). Over the past decade, significant advances have been made in biometric recognition. The goal of this course is to present in a coherent manner the various signal processing approaches for biometric recognition across different biometric modalities (e.g., face, fingerprint, and iris) and to understand the relative advantages and limitations of different approaches (e.g., image-domain approaches such as eigenfaces vs. frequency-domain approaches such as correlation filters). This course will provide the mathematical foundations underlying the approaches as well as descriptions of the state-of-the-art methods and results in biometric recognition. Computer-based projects will be an integral part of this course. The main topics include the following.
4 hrs. lec.