Estimation Aspects of Signal Spectral Components Using Neural Networks

  • Viorel Nicolau
  • Mihaela Andrei
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 195)


Many neural network models have been mathematically demonstrated to be universal approximators. For accurate function approximation, the number of samples in the training data set must be high enough to cover the entire input data space. But this number increases exponentially with the dimension of the input space, increasing the space- and time-complexity of the learning process. Hence, the neural function approximation is a complex task for problems with high dimension of the input space, like those based on signal spectral analysis. In this paper, some aspects of neural estimation of signal spectral components are discussed. The goal is to find a feed-forward neural network (FFNN) model for estimating spectral components of a signal, with computational complexity comparable with Fast Fourier Transform (FFT) algorithm, but easier to implement in hardware. Different FFNN architectures, with different data sets and training conditions, are analyzed. A butterfly-like FFNN (BFFNN) was proposed, which has much less weight connections and better performance than FFNN.


Feed-forward neural networks neural estimator spectral components signal processing FFT algorithm 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Barron, A.R.: Approximation and estimation bounds for artificial neural networks. Mach. Learn. 14(1), 115–133 (1994)MATHGoogle Scholar
  2. 2.
    Baum, B., Haussler, D.: What size net gives valid generalization? Neural Computation 1, 151–160 (1989)CrossRefGoogle Scholar
  3. 3.
    Blum, A., Rivest, R.L.: Training a 3-node neural network is NP-complete. In: Proc. of Computational Learning Theory Conference (COLT), pp. 9–18 (1988)Google Scholar
  4. 4.
    Hornik, K., Stinchcombe, K.M., White, H.: Multilayer feedforward networks are universal approximators. Neural Networks 2, 359–366 (1989)CrossRefGoogle Scholar
  5. 5.
    Hornik, K.: Approximation capabilities of multilayer feedforward networks. Neural Networks 4, 251–257 (1991)CrossRefGoogle Scholar
  6. 6.
    Huang, S.C., Huang, Y.F.: Bounds on the number of hidden neurons in multilayer perceptrons. IEEE Transactions on Neural Networks 2(1), 47–55 (1991)CrossRefGoogle Scholar
  7. 7.
    Hush, D.R., Horne, B.G.: Progress in Supervised Neural Networks. IEEE Signal processing Magazine 10(1), 8–39 (1993), doi:10.1109/79.180705CrossRefGoogle Scholar
  8. 8.
    Ilin, R., Kozma, R., Wetbos, P.J.: Beyond Feedforward Models Trained by Backpropagation: A Practical Training Tool for a More Efficient Universal Approximator. IEEE Trans. on Neural Networks 19(6), 929–937 (2008)CrossRefGoogle Scholar
  9. 9.
    Lippmann, R.P.: An Introduction to Computing with Neural Nets. IEEE ASSP Magazine 4, 4–22 (1987), doi:10.1109/MASSP.1987.1165576CrossRefGoogle Scholar
  10. 10.
    Kainen, P.C., Kurkova, V., Sanguineti, M.: Dependence of Computational Models on Input Dimension: Tractability of Approximation and Optimization Tasks. IEEE Trans. on Information Theory 58(2), 1203–1214 (2012)MathSciNetCrossRefGoogle Scholar
  11. 11.
    Keerthipala, W.W.L.: Low Tah Chong, and Tham Chong Leong Artificial neural network model for analysis of power system harmonics. In: Proc. of IEEE Int. Conf. on Neural Networks, vol. 2, pp. 905–910 (1995)Google Scholar
  12. 12.
    Leshno, M., Lin, V.Y., Pinkus, A., Schocken, S.: Multilayer feedforward networks with a nonpolynomial activation function can approximate any function. Neural Networks 6, 861–867 (1993)CrossRefGoogle Scholar
  13. 13.
    Leung, H., Haykin, S.: Rational function neural network. Neural Comput. 5, 928–938 (1993)CrossRefGoogle Scholar
  14. 14.
    Nakayama, K., Kaneda, Y., Hirano, A.: A Brain Computer Interface Based on FFT and Multilayer Neural Network. Feature Extraction and Generalization. In: Proc. of Int. Symp. on Intelligent Signal Processing and Comm. Systems (ISPACS 2007), pp. 826–829 (2007)Google Scholar
  15. 15.
    Onishi, S., Tanaka, A., Hasegawa, H., Kinoshita, K., Kishida, S.: Construction of Individual Identification System using Voice in 3-lyered Neural Networks. In: Proc. of Int. Symp. on Intelligent Signal Processing and Comm. Systems (ISPACS 2009), pp. 635–637 (2009)Google Scholar
  16. 16.
    Park, J., Sandberg, I.W.: Universal approximation using radial-basis-function networks. Neural Comput. 3, 246–257 (1991)CrossRefGoogle Scholar
  17. 17.
    Perko, M., Fajfar, I., Tuma, T., Puhan, J.: Fast Fourier Transform Computation using a Digital CNN Simulator. In: 5th IEEE Int. Workshop on Cellular Neural Networks Abd Their Applications, pp. 230–235 (1998)Google Scholar
  18. 18.
    Segee, B.E.: Using spectral techniques for improved performance in ANN. In: Proc. of IEEE Int. Conf. on Neural Networks, pp. 500–505 (1993), doi:10.1109/ICNN.1993.298608Google Scholar
  19. 19.
    Tamayo, O.M., Pineda, J.G.: Filtering and spectral processing of 1-D signals using cellular neural networks. In: Int. Symp. on Circuits and Systems (ISCAS 1996), vol. 3, pp. 76–79 (1996)Google Scholar
  20. 20.
    Wu, X., He, W., Zhang, Z., Deng, J., Li, B.: The harmonics analysis of power system based on Artificial Neural Network. In: World Automation Congress (WAC 2008), pp. 1–4 (2008)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  1. 1.Department of Electronics and Telecommunications“Dunarea de Jos” University of GalatiGalatiRomania

Personalised recommendations