Synopsis: Recognition of individuals based on physiological characteristics, such as the face, the iris or fingerprint, and behavioral traits, such as gait, is an interesting research domain, combining image and signal processing, statistics and probability theory, computer vision, deep learning, and machine learning. Biometric methods aim at the recognition of individuals in physical or logical access control application and in law enforcement and thus provide an efficient and convenient alternative to knowledge-based or token-based security systems.
The course covers applications, biometric systems and their performance, the most important modalities in today's biometric systems: fingerprint, iris, and face, classification theory, presentation attack detection, privacy, ethics, and forensic biometrics.
Aim: The goal of the course is to develop:
- An understanding of the principles used in biometric systems and how to assess performance.
- Knowledge of the most important biometric modalities (finger, iris, face).
- An understanding of the relationships between biometric systems and environmental conditions (e.g. illumination, pose variations) and their impact on performance.
- Insight into classification theory.
- An understanding of presentation attacks and methods to detect them as much as possible.
- An understanding of technical approaches to privacy, and ethics of face recognition.
- Knowledge of specific aspects of forensic biometrics in relation to general biometrics.
- Sufficient background knowledge on biometrics to read and understand scientific publications on this topic.
Learning outcomes: These are reflected in the aims described above.
Lecturers: Dr Chris Zeinstra, Dr Luuk Spreeuwers, and Prof Dr Didier Meuwly.
Examination: A few compulsory (group) assignments and a final written exam (closed-book).
Core text: Various papers from the literature