Semiparametric score level fusion: Gaussian copula approach

N. Susyanyo, C.A.J. Klaassen, Raymond N.J. Veldhuis, Lieuwe Jan Spreeuwers

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic

82 Downloads (Pure)

Abstract

Score level fusion is an appealing method for combining multi-algorithms, multi- representations, and multi-modality biometrics due to its simplicity. Often, scores are assumed to be independent, but even for dependent scores, accord- ing to the Neyman-Pearson lemma, the likelihood ratio is the optimal score level fusion if the underlying distributions are known. However, in reality, the dis- tributions have to be estimated. The common approaches are using parametric and nonparametric models. The disadvantage of the parametric method is that sometimes it is very dicult to choose the appropriate underlying distribution, while the nonparametric method is computationally expensive when the dimen- sionality increases. Therefore, it is natural to relax the distributional assumption and make the computation cheaper using a semiparametric approach. In this paper, we will discuss the semiparametric score level fusion using Gaussian copula. The theory how this method improves the recognition perfor- mance of the individual systems is presented and the performance using synthetic data will be shown. We also apply our fusion method to some public biomet- ric databases (NIST and XMVTS) and compare the thus obtained recognition performance with that of several common score level fusion rules such as sum, weighted sum, logistic regression, and Gaussian Mixture Model.
Original languageUndefined
Title of host publicationProceedings of the 36th WIC Symposium on Information Theory in the Benelux
Place of PublicationBrussels
PublisherUniversité Libre de Bruxelles
Pages26-33
Number of pages8
ISBN (Print)978-2-8052-0277-3
Publication statusPublished - 6 May 2015
Event36th WIC Symposium on Information Theory in the Benelux 2015 - Brussels, Belgium
Duration: 6 May 2015 → 7 May 2015
Conference number: 36

Publication series

Name
PublisherUniversité Libre de Bruxelles

Workshop

Workshop36th WIC Symposium on Information Theory in the Benelux 2015
CountryBelgium
CityBrussels
Period6/05/15 → 7/05/15

Keywords

  • EWI-26075
  • SCS-Safety
  • Biometric
  • IR-96056
  • Gaussian copula
  • METIS-312633
  • Fusion

Cite this

Susyanyo, N., Klaassen, C. A. J., Veldhuis, R. N. J., & Spreeuwers, L. J. (2015). Semiparametric score level fusion: Gaussian copula approach. In Proceedings of the 36th WIC Symposium on Information Theory in the Benelux (pp. 26-33). Brussels: Université Libre de Bruxelles.
Susyanyo, N. ; Klaassen, C.A.J. ; Veldhuis, Raymond N.J. ; Spreeuwers, Lieuwe Jan. / Semiparametric score level fusion: Gaussian copula approach. Proceedings of the 36th WIC Symposium on Information Theory in the Benelux. Brussels : Université Libre de Bruxelles, 2015. pp. 26-33
@inproceedings{92f607e063e442d18d378d1dd819c89e,
title = "Semiparametric score level fusion: Gaussian copula approach",
abstract = "Score level fusion is an appealing method for combining multi-algorithms, multi- representations, and multi-modality biometrics due to its simplicity. Often, scores are assumed to be independent, but even for dependent scores, accord- ing to the Neyman-Pearson lemma, the likelihood ratio is the optimal score level fusion if the underlying distributions are known. However, in reality, the dis- tributions have to be estimated. The common approaches are using parametric and nonparametric models. The disadvantage of the parametric method is that sometimes it is very dicult to choose the appropriate underlying distribution, while the nonparametric method is computationally expensive when the dimen- sionality increases. Therefore, it is natural to relax the distributional assumption and make the computation cheaper using a semiparametric approach. In this paper, we will discuss the semiparametric score level fusion using Gaussian copula. The theory how this method improves the recognition perfor- mance of the individual systems is presented and the performance using synthetic data will be shown. We also apply our fusion method to some public biomet- ric databases (NIST and XMVTS) and compare the thus obtained recognition performance with that of several common score level fusion rules such as sum, weighted sum, logistic regression, and Gaussian Mixture Model.",
keywords = "EWI-26075, SCS-Safety, Biometric, IR-96056, Gaussian copula, METIS-312633, Fusion",
author = "N. Susyanyo and C.A.J. Klaassen and Veldhuis, {Raymond N.J.} and Spreeuwers, {Lieuwe Jan}",
year = "2015",
month = "5",
day = "6",
language = "Undefined",
isbn = "978-2-8052-0277-3",
publisher = "Universit{\'e} Libre de Bruxelles",
pages = "26--33",
booktitle = "Proceedings of the 36th WIC Symposium on Information Theory in the Benelux",

}

Susyanyo, N, Klaassen, CAJ, Veldhuis, RNJ & Spreeuwers, LJ 2015, Semiparametric score level fusion: Gaussian copula approach. in Proceedings of the 36th WIC Symposium on Information Theory in the Benelux. Université Libre de Bruxelles, Brussels, pp. 26-33, 36th WIC Symposium on Information Theory in the Benelux 2015, Brussels, Belgium, 6/05/15.

Semiparametric score level fusion: Gaussian copula approach. / Susyanyo, N.; Klaassen, C.A.J.; Veldhuis, Raymond N.J.; Spreeuwers, Lieuwe Jan.

Proceedings of the 36th WIC Symposium on Information Theory in the Benelux. Brussels : Université Libre de Bruxelles, 2015. p. 26-33.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic

TY - GEN

T1 - Semiparametric score level fusion: Gaussian copula approach

AU - Susyanyo, N.

AU - Klaassen, C.A.J.

AU - Veldhuis, Raymond N.J.

AU - Spreeuwers, Lieuwe Jan

PY - 2015/5/6

Y1 - 2015/5/6

N2 - Score level fusion is an appealing method for combining multi-algorithms, multi- representations, and multi-modality biometrics due to its simplicity. Often, scores are assumed to be independent, but even for dependent scores, accord- ing to the Neyman-Pearson lemma, the likelihood ratio is the optimal score level fusion if the underlying distributions are known. However, in reality, the dis- tributions have to be estimated. The common approaches are using parametric and nonparametric models. The disadvantage of the parametric method is that sometimes it is very dicult to choose the appropriate underlying distribution, while the nonparametric method is computationally expensive when the dimen- sionality increases. Therefore, it is natural to relax the distributional assumption and make the computation cheaper using a semiparametric approach. In this paper, we will discuss the semiparametric score level fusion using Gaussian copula. The theory how this method improves the recognition perfor- mance of the individual systems is presented and the performance using synthetic data will be shown. We also apply our fusion method to some public biomet- ric databases (NIST and XMVTS) and compare the thus obtained recognition performance with that of several common score level fusion rules such as sum, weighted sum, logistic regression, and Gaussian Mixture Model.

AB - Score level fusion is an appealing method for combining multi-algorithms, multi- representations, and multi-modality biometrics due to its simplicity. Often, scores are assumed to be independent, but even for dependent scores, accord- ing to the Neyman-Pearson lemma, the likelihood ratio is the optimal score level fusion if the underlying distributions are known. However, in reality, the dis- tributions have to be estimated. The common approaches are using parametric and nonparametric models. The disadvantage of the parametric method is that sometimes it is very dicult to choose the appropriate underlying distribution, while the nonparametric method is computationally expensive when the dimen- sionality increases. Therefore, it is natural to relax the distributional assumption and make the computation cheaper using a semiparametric approach. In this paper, we will discuss the semiparametric score level fusion using Gaussian copula. The theory how this method improves the recognition perfor- mance of the individual systems is presented and the performance using synthetic data will be shown. We also apply our fusion method to some public biomet- ric databases (NIST and XMVTS) and compare the thus obtained recognition performance with that of several common score level fusion rules such as sum, weighted sum, logistic regression, and Gaussian Mixture Model.

KW - EWI-26075

KW - SCS-Safety

KW - Biometric

KW - IR-96056

KW - Gaussian copula

KW - METIS-312633

KW - Fusion

M3 - Conference contribution

SN - 978-2-8052-0277-3

SP - 26

EP - 33

BT - Proceedings of the 36th WIC Symposium on Information Theory in the Benelux

PB - Université Libre de Bruxelles

CY - Brussels

ER -

Susyanyo N, Klaassen CAJ, Veldhuis RNJ, Spreeuwers LJ. Semiparametric score level fusion: Gaussian copula approach. In Proceedings of the 36th WIC Symposium on Information Theory in the Benelux. Brussels: Université Libre de Bruxelles. 2015. p. 26-33

Access to Document