Mulitmodal Biometric Systems Munir Bandukwala Rajesh Subramanian
Introduction Biometric systems using single biometrics have to contend with Noisy data. Restricted degrees of freedom. Non universality of data. Unacceptable error rates.
Intoduction-contd… Multibiometric systems reduce these problems by providing multiple evidences of the same entity. These systems improve overall performance. Multimodal systems also make spoofing difficult. By prompting for a random set of biometric traits, such systems ensure a built in ‘liveness test’.
Problems involved A multimodal system needs a particular ‘fusion scheme’ to integrate the data from multiple modalities. Several fusion schemes are available in the literature Sum Rule [1] Decision trees [1] Linear Discriminant Analysis [1]
Problems involved-contd… Further schemes like: Learning user specific matching thresholds Weighting of individual biometric traits or A combination of the above also seem to bolster performance [2]
Problem Statement We plan to propose a fusion system that will provide an integration scheme for Fingerprint, Hand Geometry and Signature. We propose to find a scheme providing better results. The fusion will be at the matching score level.
References [1] Information Fusion in Biometrics-Arun Ross and Anil Jain [2]Learning User-specific parameters in a multibiometric system-Anil K. Jain and Arun Ross