Fusion of Dependent and Independent Biometric Information Sources
CALGARY UNIV (ALBERTA) DEPT OF ELECTRICAL AND COMPUTER ENGINEERING
Pagination or Media Count:
In this report, an overview of information fusion techniques for dependent and independent sources, specifically for biometric applications, is provided. The information fusion architecture is presented for both dependent and independent sources addressing in detail the various fusion techniques at four different levels namely raw data or signal level, feature level, decision level and multi-level integrated fusion. Furthermore, the report addresses the question of whether independent biometric sources can be fused to provide multi-modal biometric system with enhanced performance. The report shows that even when the sources are independent, the performance of a multi-modal biometric system can be better than that of a biometric system based on single source. The performance is measured in terms of total false accept rate FAR and false rejection rate FRR. The conditions for achieving an improved performance for the decision level fusion using AND, OK and majority voting are derived theoretically and confirmed through computer simulations.
- Information Science
- Biomedical Instrumentation and Bioengineering