Structures of Covariance Matrix in Handwritten Character Recognition

  • Šarūnas Raudys
  • Masakazu Iwamura
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3138)

Abstract

The integrated approach is a classifier established on statistical estimator and artificial neural network. This consists of preliminary data whitening transformation which provides good starting weight vector, and fast training of single layer perceptron (SLP). If sample size is extremely small in comparison with dimensionality, this approach could be ineffective. In the present paper, we consider joint utilization of structures and conventional regularization techniques of sample covariance matrices in order to improve recognition performance in very difficult case where dimensionality and sample size do not differ essentially. The techniques considered reduce a number of parameters estimated from training set. We applied our methodology to handwritten Japanese character recognition and found that combination of the integrated approach, conventional regularization and various structurization methods of covariance matrix outperform other methods including optimized Regularized Discriminant Analysis (RDA).

Keywords

Generalization Error Character Pair Sample Covariance Matrice Joint Utilization Handwritten Character Recognition 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

References

  1. 1.
    Morgera, D., Cooper, D.B.: Structurized estimation: Sample size reduction for adaptive pattern classification. IEEE Trans. Information Theory 23, 728–741 (1977)MATHCrossRefGoogle Scholar
  2. 2.
    Kligys, V.: On the classification of multivariate Markov sequences. In: Raudys, S. (ed.) Statistical Problems of Control, vol. 50, pp. 57–75. Inst. of Math. and Cyb. Press, Vilnius (1981) (in Russian)Google Scholar
  3. 3.
    Landgrebe, D.A.: The development of a spectral-spatial classifier for earth observational data. Pattern Recognition 12, 175–185 (1980)CrossRefGoogle Scholar
  4. 4.
    Morgera, D.: Linear, structured covariance estimation: An application to pattern classification for remote sensing. Pattern Recognition Letters 4(1), 1–7 (1986)MATHCrossRefGoogle Scholar
  5. 5.
    Palubinskas, G.: Spatial image recognition. In: Raudys, S. (ed.) Statistical Problems of Control, vol. 74, pp. 104–113. Inst. of Math. and Cyb. Press, Vilnius (1986) (in Russian)Google Scholar
  6. 6.
    Palubinskas, G.: A comparative study of decision making algorithms in images modeled by Gaussian random fields. Int. J. of Pattern Recognition and Artificial Intelligence 2(4), 621–639 (1988)CrossRefGoogle Scholar
  7. 7.
    Palubinskas, G.: A review of spatial image recognition methods. In: Raudys, S. (ed.) Statistical Problems of Control, vol. 93, pp. 215–231. Inst. of Math. and Cyb. Press, Vilnius (1990) (in Russian)Google Scholar
  8. 8.
    Raudys, S., Saudargiene, A.: Structures of the covariance matrices in the classifier design. In: Amin, A., Pudil, P., Dori, D. (eds.) SPR 1998 and SSPR 1998. LNCS, vol. 1451, pp. 583–592. Springer, Heidelberg (1998)CrossRefGoogle Scholar
  9. 9.
    Raudys, S., Saudargiene, A.: Tree type dependency model and sample size - dimensionality properties. IEEE Trans. on Pattern Analysis and Machine Intelligence 23(2), 233–239 (2001)CrossRefGoogle Scholar
  10. 10.
    Raudys, S.: Statistical and Neural Classifiers: An integrated approach to design. Springer, NY (2001)MATHGoogle Scholar
  11. 11.
    Raudys, S., Amari, S.: Effect of initial values in simple perception. In: Proceedings 1998 IEEE World Congress on Computational Intelligence, IJCNN 1998, pp. 1530–1535 (1998)Google Scholar
  12. 12.
    Omachi, S., Sun, F., Aso, H.: A new approximation method of the quadratic discriminant function. In: Amin, A., Pudil, P., Ferri, F., Iñesta, J.M. (eds.) SPR 2000 and SSPR 2000. LNCS, vol. 1876, pp. 601–610. Springer, Heidelberg (2000)CrossRefGoogle Scholar
  13. 13.
    Sun, F., Omachi, S., Kato, N., Aso, H., Kono, S., Takagi, T.: Two-stage computational cost reduction algorithm based on Mahalanobis distance approximations. In: Proceedings 15th Int. Conf. on Pattern Recognition (ICPR 2000), vol. 2, pp. 700–703. IEEE Press, Los Alamitos (2000)CrossRefGoogle Scholar
  14. 14.
    Raudys, S.: Scaled rotation regularization. Pattern Recognition 33, 1989–1998 (2000)MATHCrossRefGoogle Scholar
  15. 15.
    Sun, N., Uchiyama, Y., Ichimura, H., Aso, H., Kimura, M.: Intelligent recognition of characters using associative matching technique. In: Proc. Pacific Rim Int’l Conf. Artificial Intelligence (PRICAI 1990), pp. 546–551 (1990)Google Scholar
  16. 16.
    Yamada, H., Yamamoto, K., Saito, T.: A nonlinear normalization method for handprinted kanji character recognition - line density equalization. Pattern Recognition 23(9), 1023–1029 (1990)CrossRefGoogle Scholar
  17. 17.
    Raudys, S., Iwamura, M.: Multiple classifiers system for reducing influences of atypical observations. In: Roli, F., Kittler, J., Windeatt, T. (eds.) MCS 2004. LNCS, vol. 3077, pp. 233–242. Springer, Heidelberg (2004)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Šarūnas Raudys
    • 1
  • Masakazu Iwamura
    • 2
  1. 1.Vilnius Gediminas Technical UniversityVilniusLithuania
  2. 2.Tohoku UniversitySendaiJapan

Personalised recommendations