|
Download: |
Multibiometrics: Information Fusion in BiometricsProf. Arun Ross
Multibiometrics refers to the consolidation of multiple sources of biometric evidence in order to establish the identity of an individual. These sources of information may correspond to different biometric traits (e.g., fingerprint + voice), sensors (e.g., 2D + 3D face cameras), feature extraction and matching techniques (e.g., minutiae-based + texture-based fingerprint matchers), or instances (e.g., left and right iris). In this lecture, we will introduce various techniques that have been proposed in the literature to perform information fusion at the data level, feature level, score level, rank level, and decision level. Further, we will discuss concepts in user-specific fusion, dynamic fusion and sparse representation, where extensive training data is used to generate robust fusion rules. We will also present indexing techniques that can be used to quickly locate an identity in a large multibiometric database. Finally, we will discuss the fusion of soft biometric traits with primary biometric data in order to deduce identities in non-ideal unconstrained environments. The goal of this lecture is to introduce the audience to various fusion methods and to discuss biometric and forensic applications that benefit from them.
|
 |
