Spectrum analysis of insect-damaged wheat BPE signal based on CEEMD

Abstract

In order to effectively carry out the work of reducing losses in grain storage, it is urgently necessary to explore a fast and safe detection method for insect-damaged wheat. This work adopts a new method for extracting features. The frequency characteristics of biophoton emission (BPE) signals of insect-damaged wheat and normal wheat are obtained through Complementary Ensemble Empirical Mode Decomposition (CEEMD). Specifically, CEEMD is used to decompose the BPE signal into a series of Intrinsic Mode Functions (IMF). A Hilbert transform is then performed on the IMFs to obtain a Hilbert spectrum and Hilbert marginal spectrum. The spectral gravity frequency (SGF), spectral edge frequency (SEF), spectral amplitude proportions of different frequency bands, and energy proportion of each IMF of the BPE signals of insect-damaged wheat and normal wheat are calculated. Experiments show that the intensity of the BPE signal is relatively weak and that the BPE signal is a low frequency signal (less than 0.1Hz). The parameters of the BPE signals mentioned above give a detail spectrum analysis and show differences between insect-damaged wheat and normal wheat. We further use BP neural network to classify the wheat. The results show that this method is feasible for detecting insect-damaged wheat.