BPNN Based Processing for End Effects of HHT

This paper describes a method of signal process applied on an end effects of Hilbert-Huang transform (HHT) to provide an improvement in the reality of spectrum. The method is based on back-propagation network (BPN). To improve the effect, the end extension of the original signal is obtained by back-propagation network. A full waveform including origin and its extension is decomposed by using empirical mode decomposition (EMD) to obtain intrinsic mode functions (IMFs) of the waveform. Then, the Hilbert transform (HT) is applied to the IMFs to obtain the Hilbert spectrum of the waveform. As a result, the method is superiority of the processing of end effect of HHT to obtain the real frequency spectrum of signals.

Authors:

• Chun-Yao Lee
• Yao-chen Lee

Keywords:

• Neural network
• back-propagation network
• Hilbert-Huang transform

References:

[1] Huang, N., Z. Shen, S. Long, M. Wu, H. Shih, Q. Zheng, N. Yen, C.C.
Tung and H. Liu, " The empirical mode decomposition and the Hilbert
spectrum for nonlinear, nonstationary time series analysis," 1998 Proc. R.
Soc. Lond. Vol. A454, pp. 903-995.
[2] Y.Deng, W.Wang, C.Qian, Z.Wang, D.Dai, "Boundary-processingtechnique
in EMD method and Hilbert transform," 2001 Chinese Science
Bulletin Vol. 46, pp. 954-960.
[3] D.E.Rumelhart, G.E.Hinton, and R.J.Willia, "Learning representations by
back-propagation errors," 1986 Nature Vol. 323, pp. 533-536.
[4] Plaut,D. C., Nowlan,S. J. and Hinton,G. E., "Experiments on Learning by
Back Propagation," 1986 Technical Report CMU-CS-86-126.