Advertisement

Adaptive Score Functions for Maximum Likelihood ICA

  • Juha Karvanen
  • Jan Eriksson
  • Visa Koivunen
Article

Abstract

We propose Blind Source Separation (BSS) techniques that are applicable to a wide class of source distributions that may be skewed or symmetric and may even have zero kurtosis. Skewed distributions are encountered in many important application areas such as communications and biomedical signal processing. The methods stem from maximum likelihood approach. Powerful parametric models based on the Extended Generalized Lambda Distribution (EGLD) and the Pearson system are employed in finding the score function. Model parameters are adaptively estimated using conventional moments or linear combinations of order statistics (L-moments). The developed methods are compared with the existing methods quantitatively. Simulation examples demonstrate the good performance of the proposed methods in the cases where the standard Independent Component Analysis (ICA) methods perform poorly.

Blind Separation score function Generalized Lambda Distribution Pearson system L-moments 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    S. Choi, A. Cichocki, and S. Amari, “Flexible Independent Component Analysis,” in Neural Networks for Signal Processing VIII, 1998, Proceedings of the 1998 IEEE Signal Processing Society Workshop, 1998, pp. 83-92.Google Scholar
  2. 2.
    J. Cao and N. Murata, “A Stable and Robust ICA Algorithm Based on t-distribution and Generalized Gaussian Distribution Models,” in Neural Networks for Signal Processing IX, 1999, The IEEE 1999 Proceedings, 1999, pp. 283-292.Google Scholar
  3. 3.
    A. Hyvärinen, “Fast and Robust Fixed-Point Algorithms for Independent Component Analysis,” IEEE Transactions on Neural Networks, vol. 10,no. 3, 1999, pp. 626-634.CrossRefGoogle Scholar
  4. 4.
    J.F. Cardoso, “Blind Signal Separation: Statistical Principles,” Proceedings of the IEEE, vol. 86,no. 10, 1998, pp. 2009-2025.CrossRefGoogle Scholar
  5. 5.
    J.F. Cardoso, “Infomax and Maximum Likelihood for Blind Signal Processing,” IEEE Signal Processing Letters, vol. 4,no. 4, 1997.Google Scholar
  6. 6.
    T.-W. Lee, Independent Component Analysis: Theory and applications, Boston: Kluwer Academic Publishers, 1998.zbMATHGoogle Scholar
  7. 7.
    S.-I. Amari and A. Cichocki, “Adaptive Blind Signal Processing—Neural Network Approaches,” Proceedings of the IEEE, vol. 86,no. 10, 1998, pp. 2026-2048.CrossRefGoogle Scholar
  8. 8.
    D.T. Pham, “Blind Separation of Instantaneous Mixture of Sources via an Independent Component Analysis,” IEEE Transactions on Signal Processing, vol. 44,no. 11, 1996, pp. 2768-2779.CrossRefGoogle Scholar
  9. 9.
    J. Eriksson, J. Karvanen, and V. Koivunen, “Source Distribution Adaptive Maximum Likelihood Estimation of ICA Model,” in ICA2000, Proceedings of the Second International Workshop on Independent Component Analysis and Blind Signal Separation, 2000, pp. 227-232.Google Scholar
  10. 10.
    J. Karvanen, J. Eriksson, and V. Koivunen, “Pearson System Based Method for Blind Separation,” in ICA2000, Proceedings of the Second International Workshop on Independent Component Analysis and Blind Signal Separation, 2000, pp. 585-590.Google Scholar
  11. 11.
    Z.A. Karian, E.J. Dudewicz, and P. McDonald, “The Extended Generalized Lambda Distribution System for Fitting Distributions to Data: History, Completion of Theory, Tables, Applications, the “Final Word” on Moment Fits,” Communications in Statistics: Simulation and Computation, vol. 25,no. 3, 1996, pp. 611-642.MathSciNetCrossRefzbMATHGoogle Scholar
  12. 12.
    A. Stuart and J.K. Ord, Kendall's Advanced Theory of Statistics: Distribution Theory, vol. 1, 6th edn., London: Edward Arnold, 1994.Google Scholar
  13. 13.
    J. Hosking, “L-Moments: Analysis and Estimation of Distributions Using Linear Combinations of Order Statistics,” Journal of Royal Statistical Society B, vol. 52,no. 1, 1990, pp. 105-124.MathSciNetzbMATHGoogle Scholar
  14. 14.
    Z.A. Karian and E.J. Dudewicz, “The Extended Generalized Lambda Distribution (EGLD) System for Fitting Distributions to Data with Moments, II: Tables,” American Journal of Mathematical and Management Sciences, 1996.Google Scholar
  15. 15.
    J. Eriksson, J. Karvanen, and V. Koivunen, EGLD-ICA Matlab Code Available at http://wooster.hut.fi/statsp/publications.html, 2000.Google Scholar
  16. 16.
    S.-I. Amari, “Natural Gradient Works Efficiently in Learning,” Neural Computation, vol. 10, 1998, pp. 251-276.CrossRefGoogle Scholar
  17. 17.
    A. Hyvärinen, “The Fixed-Point Algorithm and Maximum Likelihood Estimation for Independent Component Analysis,” Neural Processing Letters, vol. 10,no. 1, 1999, pp. 1-5.CrossRefGoogle Scholar
  18. 18.
    J. Karvanen, J. Eriksson, and V. Koivunen, Pearson-ICA Matlab Code Available at http://wooster.hut.fi/statsp/publications.html, 2000.Google Scholar
  19. 19.
    G.W. Steward and J.-G. Sun, Matrix Perturbation Theory, Boston: Academic Press, 1990.Google Scholar
  20. 20.
    S.-I. Amari, A. Cichocki, and H. Yang, “A New Learning Algorithm for Blind Signal Separation,” in Advances in Neural Information Processing Systems, vol. 8, Cambridge MA: MIT Press, 1996, pp. 757-763.Google Scholar
  21. 21.
    J.F. Cardoso, JADE Matlab Code with References Available at http://sig.enst.fr:80/~cardoso/.Google Scholar
  22. 22.
    T.-W. Lee, M. Girolami, and T. Sejnowski, Infomax Matlab Code with References Available at http://www.cnl.salk.edu/~scott/ica-download-form.html.Google Scholar
  23. 23.
    A. Hyvärinen, FastICA Matlab Code with References Available at http://www.cis.hut.fi/projects/ica/fastica/, 1998.Google Scholar
  24. 24.
    S. Cruces, L. Castedo, and A. Cichocki, “Novel Blind Source Separation Algorithms Using Cumulants,” in Proc. of ICASSP, 2000, pp. 3152-3155.Google Scholar

Copyright information

© Kluwer Academic Publishers 2002

Authors and Affiliations

  • Juha Karvanen
    • 1
  • Jan Eriksson
    • 1
  • Visa Koivunen
    • 1
  1. 1.Signal Processing LaboratoryHelsinki University of TechnologyHUTFinland

Personalised recommendations