SPECTRAL MINUTIAE REPRESENTATIONS

,The term biometrics refers to the technologies that measure and analyze human intrinsic physical or behavioral characteristics for authenticating individuals. Nowadays, biometric technology is increasingly deployed in civil and commercial applications. The growing use of biometrics is raising security and privacy concerns. Storing biometric data, known as biometric templates, in a database leads to several privacy risks such as identity fraud and cross matching. A solution is to apply biometric template protection techniques, which aim to make it impossible to recover the biometric data from the templates. The goal of our research is to combine biometric systems with template protection. Aimed at fingerprint recognition, this thesis introduces the Spectral Minutiae Representation method, which enables the combination of a minutiae-based fingerprint recognition system with template protection schemes based on fuzzy commitment or helper data schemes. In this thesis, three spectral minutiae representation methods have been proposed: the location-based spectral minutiae representation (SML the orientation-based spectral minutiae representation (SMO) and the complex spectral minutiae representation (SMC). From the experiments shown in this thesis, SMC achieved the best results. Based on the spectral minutiae features, this thesis further presented contributions in three research directions. First, this thesis recommends several ways to enhance the recognition performance of SMC. Second, with regard to feature reduction, this thesis introduced two feature reduction methods, Column-PCA (CPCA) and Line-DFT (LDFT). Third, with regard to quantization, this thesis introduced the Spectral Bits and Phase Bits representations. The spectral minutiae representation scheme proposed in this thesis enables the combination of fingerprint recognition systems with template protection based on the helper data scheme. Furthermore, this scheme allows for a fast minutiae comparison, which renders this scheme suitable as a pre-selector for a large-scale fingerprint identification system, thus significantly reducing the time to perform matching. The binary spectral minutiae representation achieved an equal error rate of less than 1% on the FVC2000-DB2 database when applying multi-sample enrolment. The fast comparison speed together with the promising recognition performance makes this spectral minutiae scheme very applicable for real time applications.