Advertisement

Robust Fusion: Extreme Value Theory for Recognition Score Normalization

  • Walter Scheirer
  • Anderson Rocha
  • Ross Micheals
  • Terrance Boult
Part of the Lecture Notes in Computer Science book series (LNCS, volume 6313)

Abstract

Recognition problems in computer vision often benefit from a fusion of different algorithms and/or sensors, with score level fusion being among the most widely used fusion approaches. Choosing an appropriate score normalization technique before fusion is a fundamentally difficult problem because of the disparate nature of the underlying distributions of scores for different sources of data. Further complications are introduced when one or more fusion inputs outright fail or have adversarial inputs, which we find in the fields of biometrics and forgery detection. Ideally a score normalization should be robust to model assumptions, modeling errors, and parameter estimation errors, as well as robust to algorithm failure. In this paper, we introduce the w-score, a new technique for robust recognition score normalization. We do not assume a match or non-match distribution, but instead suggest that the top scores of a recognition system’s non-match scores follow the statistical Extreme Value Theory, and show how to use that to provide consistent robust normalization with a strong statistical basis.

References

  1. 1.
    Huber, P.: Robust Statistics. Wiley, New York (1981)zbMATHCrossRefGoogle Scholar
  2. 2.
    Li, W., Gao, X., Boult, T.E.: Predicting Biometric System Failure. In: CIHSPS (2005)Google Scholar
  3. 3.
    Wang, P., Ji, Q., Wayman, J.: Modeling and Predicting Face Recognition System Performance Based on Analysis of Similarity Scores. IEEE TPAMI 29, 665–670 (2007)Google Scholar
  4. 4.
    Shakhnarovich, G., Fisher, J., Darrell, T.: Face Recognition From Long-term Observations. In: Heyden, A., Sparr, G., Nielsen, M., Johansen, P. (eds.) ECCV 2002. LNCS, vol. 2352, pp. 851–868. Springer, Heidelberg (2002)CrossRefGoogle Scholar
  5. 5.
    Jain, A., Nandakumar, K., Ross, A.: Score Normalization in Multimodal Biometric Systems. Pattern Recognition 38, 2270–2285 (2005)CrossRefGoogle Scholar
  6. 6.
    Hampel, F., Rousseeuw, P., Ronchetti, E., Stahel, W.: Robust Statistics: The Approach Based on Influence Functions. Wiley, New York (1986)zbMATHGoogle Scholar
  7. 7.
    Poh, N., Bengio, S.: How Do Correlation and Variance of Base Classifiers Affect Fusion in Biometric Authentication Tasks? IEEE. TSP 53, 4384–4396 (2005)MathSciNetGoogle Scholar
  8. 8.
    Poh, N., Kittler, J.: Incorporating Variation of Model-specific Score Distribution in Speaker Verification Systems. IEEE. TASLP 16, 594–606 (2008)Google Scholar
  9. 9.
    Poh, N., Bourlai, N., Kittler, J.: Benchmarking Quality-Dependent and Cost-Sensitive Score-Level Multimodal Biometric Fusion Algorithms. IEEE. TIFS 4, 849–866 (2009)Google Scholar
  10. 10.
    Poh, N., Bourlai, T., Kittler, J.: A Multimodal Biometric Test Bed for Quality-dependent, Cost-sensitive and Client-specific Score-level Fusion Algorithms. Pattern Recognition 43, 1094–1105 (2010)zbMATHCrossRefGoogle Scholar
  11. 11.
    Shi, Z., Kiefer, F., Schneider, J., Govindaraju, V.: Modeling Biometric Systems Using the General Pareto Dstribution (GPD). In: SPIE, vol. 6944 (2008)Google Scholar
  12. 12.
    Grother, P., Phillips, P.J.: Models of Large Population Recognition Performance. In: IEEE CVPR, pp. 68–75 (2004)Google Scholar
  13. 13.
    Broadwater, J., Chellappa, R.: Adaptive Threshold Estimation Via Extreme Value Theory. IEEE TSP (2009) (to appear)Google Scholar
  14. 14.
    Kotz, S., Nadarajah, S.: Extreme Value Distributions: Theory and Applications, 1st edn. World Scientific, Singapore (2001)Google Scholar
  15. 15.
    Gumbel, E.: Statistical Theory of Extreme Values and Some Practical Applications. In: Number National Bureau of Standards Applied Mathematics in 33, U.S. GPO, Washington, D.C (1954)Google Scholar
  16. 16.
    NIST: NIST/SEMATECH Handbook of Statistical Methods. 33. U.S. GPO (2008)Google Scholar
  17. 17.
    NIST: Biometric Scores Set (2004), www.itl.nist.gov/iad/894.03/biometricscores
  18. 18.
    Datta, R., Joshi, D., Wang, J.: Image retrieval: Ideas, influences, and trends of the new age. ACM CSUR 40, 1–77 (2008)CrossRefGoogle Scholar
  19. 19.
    Stehling, R., Nascimento, M., Falcão, A.: A compact and efficient image retrieval approach based on border/interior pixel classification. In: CIKM, pp. 102–109 (2002)Google Scholar
  20. 20.
    Jegou, H., Douze, M., Schmid, C.: Hamming embedding and weak geometry consistency for large scale image search. In: Forsyth, D., Torr, P., Zisserman, A. (eds.) ECCV 2008, Part I. LNCS, vol. 5302, pp. 304–317. Springer, Heidelberg (2008)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Walter Scheirer
    • 1
  • Anderson Rocha
    • 2
  • Ross Micheals
    • 3
  • Terrance Boult
    • 1
  1. 1.University of Colorado at Colorado Springs & Securics, Inc. 
  2. 2.Institute of ComputingUniversity of Campinas 
  3. 3.National Institute of Standards and Technology 

Personalised recommendations