In this study, we attempt to gain an understanding of whether or not a change in performance of a biometric system is subject-dependent. We will not attempt to explain the cause of performance change; we know that it is not only age-related, but is also dependent on biometric sample quality and habituation. What will become clear is that measuring the change in biometric performance is challenging, let alone explaining the causes of this change. However, we would challenge the common assertion that the performance of a biometric system systematically degrades overtime.

Chapter Contents:

• 10.1 Introduction
• 10.2 User-specific performance characterisation
• 10.3 Our framework: a homomorphic users grouping algorithm
• 10.4 Experiment setup
• 10.4.1 Database and experimental protocol
• 10.4.2 Face classifiers
• 10.4.3 Speech classifier
• 10.4.4 Fusion classifier
• 10.5 Results
• 10.5.1 Results I: Model fitting
• 10.5.2 Results II: Partitioned subjects
• 10.6 Conclusions
• Appendix A: The face classifier
• A.1 Local binary pattern
• A.2 Local phase quantisation pattern
• A.3 Multiscale pattern histogram
• A.4 Image frame matching
• A.5 Video face matching
• Acknowledgement
• References

Inspec keywords: biometrics (access control); human factors

Other keywords: biometric sample quality; multimodal perspective; biometric system; biometric performance change

Subjects: Ergonomic aspects of computing