Managing ageing effects in biometric systems

• Main Author: Erbilek, Meryem
• Published: University of Kent 2013
• Subjects: 571.878
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.596077

Biometric systems deal with the problem of determining or authenticating the identity of individuals based on measurements of their physiological or behavioural characteristics. However, these characteristics are likely to change with the natural ageing process (passage of time) and, as a result, developing biometric applications for long-term use becomes a particularly challenging task. Thus, increasingly, an understanding of the ageing process is becoming an important issue in terms of ensuring reliability in the face of changing data scenarios in biometric systems. In fact, both physical ageing (the effects of ageing across the whole ageing cycle on the biometric measurements used in the identification process) and template ageing (the effects of ageing across the time elapsed between enrolment and authentication) are important in a pmcticaJ context, and while these are obviously related phenomena, they are not the same thing, since template ageing also depends to some extent on the physical age of the subject within the relevant time frame. This thesis will describe a study to explore some important issues related to physical ageing issues of particular importance in biometric systems, in order to provide an improved understanding of the ageing problem which might be able to have a positive influence on the design, deployment and management of ageing issues in future biometric systems. We will explore and present quantitatively the results of a detaiJed investigation into the physical effects of ageing, will discuss the relationships between physical ageing and interrelated physical factors which have a bearing on how the impact of ageing can most effectively be investigated and understood, and will show how these factors can be manipulated in order to guide practical implementation towards achieving more reliable performance. We study two established and widely used biometric modalities, iris and signature, to provide a practical environment for experimentation and analysis.