On the Optimality of the Simple Bayesian Classifier under Zero-One Loss
Machine Learning - Special issue on learning with probabilistic representations
Integrating Faces and Fingerprints for Personal Identification
IEEE Transactions on Pattern Analysis and Machine Intelligence
Information fusion in biometrics
Pattern Recognition Letters - Special issue: Audio- and video-based biometric person authentication (AVBPA 2001)
Comparison of Face Verification Results on the XM2VTS Database
ICPR '00 Proceedings of the International Conference on Pattern Recognition - Volume 4
The BANCA database and evaluation protocol
AVBPA'03 Proceedings of the 4th international conference on Audio- and video-based biometric person authentication
Automatic configuration for a biometrics-based physical access control system
IWBRS'05 Proceedings of the 2005 international conference on Advances in Biometric Person Authentication
Hi-index | 0.00 |
Building access control represents an important application for biometric verification but often requires greater accuracy than can be provided by a single verifier. Even as algorithms continue to improve, poor samples and environmental factors will continue to impact performance in the building environment. We aim to improve verification accuracy by combining decisions from multiple verifiers spread throughout a building. In particular, we combine verifiers along the path traced out by each subject. When combining these decisions, the assumption of conditional independence simplifies implementation but can potentially lead to suboptimal performance. Through empirical evaluation of two related algorithms, we show that the assumption of conditional independence does not significantly impact verification accuracy. We argue that such a small reduction in accuracy can be attributed to the relative accuracy of each verifier. As a result, decisions can be combined using low complexity fusion rules without concerns of degraded accuracy.