Improving fusion with one-class classification and boosting in multimodal biometric authentication

Quang Duc Tran, Panos Liatsis

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations


Class imbalance poses serious difficulties to most standard two-class classifiers, when applied in performing classification in the context of multimodal biometric authentication. In this paper, we propose a system, which exploits the natural capabilities of one-class classifiers in conjunction with the so-called Real AdaBoost to handle the class imbalance problem in biometric systems. Particularly, we propose a decision rule for the fusion of one-class classifiers to effectively use the training data from both classes. By treating this decision rule as the base classifier, the Real AdaBoost is then employed to further improve its performance. An important feature of the proposed system is that it trains the base classifiers with different parameter settings. Hence, it is able to reduce the number of parameters, which are normally set by the user. An empirical evaluation, carried out on the BioSecure DS2 database, demonstrates that the proposed system can achieve a relative performance improvement of 5%, 13%, and 14% as compared to other state-of-the-art techniques, namely the sum of scores, likelihood ratio based score fusion, and support vector machines.

Original languageBritish English
Title of host publicationIntelligent Computing in Bioinformatics - 10th International Conference, ICIC 2014, Proceedings
PublisherSpringer Verlag
Number of pages7
ISBN (Print)9783319093291
StatePublished - 2014
Event10th International Conference on Intelligent Computing, ICIC 2014 - Taiyuan, China
Duration: 3 Aug 20146 Aug 2014

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8590 LNBI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference10th International Conference on Intelligent Computing, ICIC 2014


  • boosting
  • Class imbalance
  • one-class classification
  • Real AdaBoost


Dive into the research topics of 'Improving fusion with one-class classification and boosting in multimodal biometric authentication'. Together they form a unique fingerprint.

Cite this