Maximizing area under ROC curve for biometric scores fusion

Research output: Contribution to journalArticle

38 Citations (Scopus)

Abstract

The receiver operating characteristics (ROC) curve has been extensively used for performance evaluation in multimodal biometrics fusion. However, the processes of fusion classifier design and the final ROC performance evaluation are usually conducted separately. This has been inevitable because the ROC, when taken from the error counting point of view, does not have a well-posed structure linking to the fusion classifier of interest. In this work, we propose to optimize the ROC performance directly according to the fusion classifier design. The area under the ROC curve (AUC) will be used as the optimization objective since it provides a good representation of the ROC performance. Due to the piecewise cumulative structure of the AUC, a smooth approximate formulation is proposed. This enables a direct optimization of the AUC with respect to the classifier parameters. When a fusion classifier has linear parameters, computation of the solution to optimize a quadratic AUC approximation is surprisingly simple and yet effective. Our empirical experiments on biometrics fusion show strong evidences regarding the potential of the proposed method.

Original languageEnglish
Pages (from-to)3373-3392
Number of pages20
JournalPattern Recognition
Volume41
Issue number11
DOIs
Publication statusPublished - 2008 Nov 1

Fingerprint

Biometrics
Fusion reactions
Classifiers
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

@article{b4d497ddc33c4be08629bb789c9dfc29,
title = "Maximizing area under ROC curve for biometric scores fusion",
abstract = "The receiver operating characteristics (ROC) curve has been extensively used for performance evaluation in multimodal biometrics fusion. However, the processes of fusion classifier design and the final ROC performance evaluation are usually conducted separately. This has been inevitable because the ROC, when taken from the error counting point of view, does not have a well-posed structure linking to the fusion classifier of interest. In this work, we propose to optimize the ROC performance directly according to the fusion classifier design. The area under the ROC curve (AUC) will be used as the optimization objective since it provides a good representation of the ROC performance. Due to the piecewise cumulative structure of the AUC, a smooth approximate formulation is proposed. This enables a direct optimization of the AUC with respect to the classifier parameters. When a fusion classifier has linear parameters, computation of the solution to optimize a quadratic AUC approximation is surprisingly simple and yet effective. Our empirical experiments on biometrics fusion show strong evidences regarding the potential of the proposed method.",
author = "Toh, {Kar Ann} and Jaihie Kim and Sangyoun Lee",
year = "2008",
month = "11",
day = "1",
doi = "10.1016/j.patcog.2008.04.002",
language = "English",
volume = "41",
pages = "3373--3392",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier Limited",
number = "11",

}

Maximizing area under ROC curve for biometric scores fusion. / Toh, Kar Ann; Kim, Jaihie; Lee, Sangyoun.

In: Pattern Recognition, Vol. 41, No. 11, 01.11.2008, p. 3373-3392.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Maximizing area under ROC curve for biometric scores fusion

AU - Toh, Kar Ann

AU - Kim, Jaihie

AU - Lee, Sangyoun

PY - 2008/11/1

Y1 - 2008/11/1

N2 - The receiver operating characteristics (ROC) curve has been extensively used for performance evaluation in multimodal biometrics fusion. However, the processes of fusion classifier design and the final ROC performance evaluation are usually conducted separately. This has been inevitable because the ROC, when taken from the error counting point of view, does not have a well-posed structure linking to the fusion classifier of interest. In this work, we propose to optimize the ROC performance directly according to the fusion classifier design. The area under the ROC curve (AUC) will be used as the optimization objective since it provides a good representation of the ROC performance. Due to the piecewise cumulative structure of the AUC, a smooth approximate formulation is proposed. This enables a direct optimization of the AUC with respect to the classifier parameters. When a fusion classifier has linear parameters, computation of the solution to optimize a quadratic AUC approximation is surprisingly simple and yet effective. Our empirical experiments on biometrics fusion show strong evidences regarding the potential of the proposed method.

AB - The receiver operating characteristics (ROC) curve has been extensively used for performance evaluation in multimodal biometrics fusion. However, the processes of fusion classifier design and the final ROC performance evaluation are usually conducted separately. This has been inevitable because the ROC, when taken from the error counting point of view, does not have a well-posed structure linking to the fusion classifier of interest. In this work, we propose to optimize the ROC performance directly according to the fusion classifier design. The area under the ROC curve (AUC) will be used as the optimization objective since it provides a good representation of the ROC performance. Due to the piecewise cumulative structure of the AUC, a smooth approximate formulation is proposed. This enables a direct optimization of the AUC with respect to the classifier parameters. When a fusion classifier has linear parameters, computation of the solution to optimize a quadratic AUC approximation is surprisingly simple and yet effective. Our empirical experiments on biometrics fusion show strong evidences regarding the potential of the proposed method.

UR - http://www.scopus.com/inward/record.url?scp=48149091075&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=48149091075&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2008.04.002

DO - 10.1016/j.patcog.2008.04.002

M3 - Article

AN - SCOPUS:48149091075

VL - 41

SP - 3373

EP - 3392

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

IS - 11

ER -