Empirical wavelet transform method for determining sub-band boundary by using kurtosis spectrum

Abstract

The invention relates to an empirical wavelet transform method for determining a sub-band boundary by using a kurtosis spectrum. Firstly, a signal is subjected to time-frequency transform; on the basis of a time-frequency transform result, partial time-frequency regions are orderly subjected to inverse transformation along a frequency axis of the time-frequency transform result in a fixed frequency step size and fixed bandwidth; a signal component of the time-frequency region is obtained; then, the kurtosis thereof is obtained; finally, a kurtosis sequence is obtained. With the center frequency of each partial time-frequency region as the abscissa and the kurtosis of the signal component thereof as the ordinate, a kurtosis spectrum of the signal is obtained; a local minimum is found out inthe kurtosis spectrum; the corresponding frequencies thereof are rearranged; the frequencies are used as a boundary frequency of the wavelet sub-band; thereby, an analytical band of the signal is segmented into a group of wavelet bands, which are mutually connected and are not overlapped with each other; on this basis, a scale filter and a wavelet filter are constructed based on the wavelet bands; according to a wavelet decomposition principle, the signal is subjected to wavelet transform; and the signal component of the signal in each sub-band is obtained.

Description

Technical field [0001] The invention belongs to the field of signal processing, and specifically relates to an empirical wavelet transform method for determining the boundary of a subband by using a kurtosis spectrum. Background technique [0002] Empirical Wavelet Transform (EWT) is a signal adaptive processing method proposed by Gilles in 2013. It combines the complete theory of wavelet analysis and the adaptiveness of empirical mode decomposition to decompose the signal into a series of frequency modulation Amplitude modulation characteristic component. During signal decomposition, the signal is Fourier transformed first, and then the local extreme amplitudes on the frequency spectrum are arranged in descending order, and the midpoint of the two adjacent local maxima is used as the boundary of the wavelet subband (starting band The boundary of is 0, the boundary of the last frequency band is the boundary value of the signal Nyquist analysis frequency band), the signal Nyquist...

AI Technical Summary

Problems solved by technology

[0003]However, in the real world, the signal usually contains noise. For the signal with low signal-to-noise ratio, the spectral peak of the noise component in the signal affects the signal Therefore, the wavelet sub-band boundary determined by the local maximum magnitude method often loses its rationality, which makes the signal components obtained by empirical wavelet transform decomposition insufficient, which affects its engineering application

Images