Combination of Experts by Classifiers in Similarity Score Spaces
DOI: 10.1007/978-3-540-89689-0_86 Conference: Structural, Syntactic, and Statistical Pattern Recognition, Joint IAPR International Workshop, SSPR & SPR 2008, Orlando, USA, December 4-6, 2008. Proceedings
The combination of different experts is largely used to improve the performance of a pattern recognition system. In the case of experts whose output is a similarity score, different methods had been developed. In this paper, the combination is performed by building a similarity score space made up of the scores produced by the experts, and training a classifier into it. Different techniques based on the use of classifiers trained on the similarity score space are proposed and compared. In particular, they are used in the framework of Dynamic Score Selection mechanisms, recently proposed by the authors. Reported results on two biometric datasets show the effectiveness of the proposed approach.
Available from: Roberto Tronci
- "Moreover, part of the AmILAB's work is focused on research activities made in collaboration with the Pattern Recognition and Applications group of the University of Cagliari 4 . From this collaboration 8 papers were published so far: 3 on dynamic combination for biometric authentication   , 2 on face spoofing over a biometric authentication system  , and 3 on content based image retrieval   . "
[Show abstract] [Hide abstract]
ABSTRACT: Multimodal biometric systems integrate information from multiple sources to improve the performance of a typical unimodal
biometric system. Among the possible information fusion approaches, those based on fusion of match scores are the most commonly
used. Recently, a framework for the optimal combination of match scores that is based on the likelihood ratio (LR) test has
been presented. It is based on the modeling of the distributions of genuine and impostor match scores as a finite Gaussian
mixture models. In this paper, we propose two strategies for improving the performance of the LR test. The first one employs
a voting strategy to circumvent the need of huge datasets for training, while the second one uses a sequential test to improve
the classification accuracy on genuine users.
Experiments on the NIST multimodal database confirmed that the proposed strategies can outperform the standard LR test, especially
when there is the need of realizing a multibiometric system that must accept no impostors.
Data provided are for informational purposes only. Although carefully collected, accuracy cannot be guaranteed. The impact factor represents a rough estimation of the journal's impact factor and does not reflect the actual current impact factor. Publisher conditions are provided by RoMEO. Differing provisions from the publisher's actual policy or licence agreement may be applicable.