Using a Generalized Linear Mixed Model to Study the Configuration Space of a PCA+LDA Human Face Recognition Algorithm

  • Geof H. Givens
  • J. Ross Beveridge
  • Bruce A. Draper
  • David Bolme
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3179)


A generalized linear mixed model is used to estimate how rank 1 recognition of human faces with a PCA+LDA algorithm is affected by the choice of distance metric, image size, PCA space dimensionality, supplemental training, and the inclusion of test subjects in the training data. Random effects for replicated training sets and for repeated measures on people were included in the model. Results indicate that variation among subjects was a dominant source of variability, and that there was moderate correlation within people. Statistically significant effects and interactions were found for all configuration factors except image size. The most significant interaction is that dhanges to the PCA+LDA configuration only improved recognition for test subjects who were included in the training data. For subjects not included in training, no configuration changes were helpful.

This study is a model for how to evaluate algorithms with large numbers of parameters. For example, by accounting for subject variation as a random effect and explicitly looking for interaction effects, we are able to discern effects that might otherwise have been masked by subject variation and interaction effects. This study is instructive for what it reveals about PCA+LDA.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Heath, M., Sarkar, S., Sanocki, T., Bowyer, K.: A robust visual method for assessing the relative performance of edge detection algorithms. PAMI 19, 1338–1359 (1997)Google Scholar
  2. 2.
    Zhao, W., Chellappa, R., Krishnaswamy, A.: Discriminant analysis of principal components for face recognition. In: Wechsler, Philips, Bruce, Fogelman-Soulie, Huang (eds.) Face Recognition: From Theory to Applications, pp. 73–85 (1998)Google Scholar
  3. 3.
    Phillips, P., Moon, H., Rizvi, S., Rauss, P.: The FERET Evaluation Methodology for Face-Recognition Algorithms. T-PAMI 22, 1090–1104 (2000)Google Scholar
  4. 4.
    FERET Database: NIST (2001),
  5. 5.
    Turk, M.A., Pentland, A.P.: Face Recognition Using Eigenfaces. In: Proc. of IEEE Conference on Computer Vision and Pattern Recognition, pp. 586–591 (1991)Google Scholar
  6. 6.
    Beveridge, J.R.: The Geometry of LDA and PCA Classifiers Illustrated with 3D Examples. Technical Report CS-01-101, Computer Science, Colorado State University (2001)Google Scholar
  7. 7.
    Kirby, M.: Dimensionality Reduction and Pattern Analysis: An Empirical Approach. Wiley, Chichester (2000)Google Scholar
  8. 8.
    Breslow, N.E., Clayton, D.G.: Approximate inference in generalized linear mixed models. J. Amer. Statist. Assoc. 8, 9–25 (1993)CrossRefGoogle Scholar
  9. 9.
    Rao, C.R.: Estimation of variance and covariance components in linear models. J. Amer. Statist. Assoc. 67, 112–115 (1972)MATHCrossRefMathSciNetGoogle Scholar
  10. 10.
    LaMotte, L.R.: Quadratic estimation of variance components. Biometrics 29, 311–330 (1973)CrossRefMathSciNetGoogle Scholar
  11. 11.
    Wolfinger, R., O’Connell, M.: Generalized linear models: a pseudo-likelihood approach. Journal of Statistical Computation and Simulation 48, 233–243 (1993)MATHCrossRefGoogle Scholar
  12. 12.
    Wolfinger, R., Tobias, R., Sall, J.: Computing gaussian likelihoods and their derivatives for general linear mixed models. SIAM Journal of Scientific Computing 15, 1294–1310 (1994)MATHCrossRefMathSciNetGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Geof H. Givens
    • 1
  • J. Ross Beveridge
    • 2
  • Bruce A. Draper
    • 2
  • David Bolme
    • 2
  1. 1.Statistics DepartmentColorado State UniversityFort CollinsUSA
  2. 2.Department of Computer ScienceColorado State UniversityFort CollinsUSA

Personalised recommendations