Indexing on Biometric Databases

Abstract

Identification is an important authentication mechanism that determines the identity of a given sample. The process is inherently compute-intensive as it needs to search through the complete database to find a suitable match with the sample. Most of the time, database size is large and is expected to grow as more and more users get registered with the system. This leads to longer identification time, and it becomes increasingly difficult to produce results for larger databases. A technique that can filter out highly dissimilar samples from being compared by producing a small candidate list having likely to match database samples could be very useful to speed up the identification. Such a technique is called indexing. The technique would have two parts: (i) the first creates a suitable index structure called indexing, and (ii) the second produces a candidate list when a query sample is provided, called retrieval. Both the methods are related and complement each other. Developing an indexing scheme for a biometric database is challenging because of inherent properties of the biometric trait and features. This chapter highlights challenges and introduces some of the indexing techniques for popular biometric traits. This chapter considers four physiological biometric traits, viz., face, fingerprint, finger-knuckle print, and iris along with a behavioral trait, viz., signature to describe the biometric indexing.