Ship Radiated Noise Modulation Feature Extraction Based on CEEMD and Wavelet Threshold Noise Reduction

Ship propeller beats have obvious amplitude modulation to the cavitation noise of its radiation. The DEMON (Detection of Envelope Modulation on Noise) spectrum analysis technology is to demodulate the high frequency components to get the low frequency modulation components. This paper studied the new method to combine the complementary ensemble empirical mode decomposition (CEEMD) with wavelet threshold to reduce the noise. After verifying the method using both simulated signal and experimental signal, it is concluded that this algorithm has good noise suppression performance and the modulation information extracted out is clearer and more comprehensive. Introduction In regard to the modulation signal detection under the condition of the strong background noise, domestic scholars have studied the extraction of the DEMON spectrum [1] by the use of modern signal processing methods such as the higher order statistics [2,3] , the time-frequency analysis [4,5], singular value analysis and so on. The methods mentioned above improved the performance of demodulation to a certain extent, but some shortcomings, such as less obvious harmonic characteristics and poor ability of noise suppression, still exist. When the target is unstable or interfered by adjacent strong target, the methods above cannot give satisfactory results. This paper presents a new method combined the CEEMD with wavelet threshold denoising to improve the performance of ship radiated noise demodulation. Method of Complementary Ensemble Empirical Mode Decomposition The empirical mode decomposition (EMD) has the adaptive band division function and the characteristics of orthogonality, completeness and adaptability, etc. All the features above make it suitable for processing non-stationary nonlinear signal [6]. When the signal is intermittent, using the EMD method may lead to the phenomenon of modal aliasing and energy leakage, making the physical meaning of the intrinsic mode function (IMF) unclear. In this regard, the document [7] proposed a method named complementary ensemble empirical mode decomposition (CEEMD) based on the EEMD (Ensemble Empirical Mode Decomposition) [8]. Taking the advantage of characteristics that white noise power spectral density is evenly distributed, it adds positive and negative pairs of white noise to the original signal repeatedly, making signal continuous at different scales. And then eliminate the effects of the auxiliary noise with less average number, making the decomposition eventually have antinoise property. This method not only solves the modal aliasing and energy leakage problems of the EMD method, but also has higher operation efficiency. CEEMD mainly includes the following steps: a. Add the target signal with a pair of Gaussian white noise which have the same amplitude and a phase angle difference of 180° to construct two new signal 1 x and 2 x . b. Execute EMD operation to the new signal 1 x and 2 x , and each signal obtains a set of IMF, in which the j -th IMF component of the i -th signal is expressed as ij C . International Industrial Informatics and Computer Engineering Conference (IIICEC 2015) © 2015. The authors Published by Atlantis Press 1763 c. Obtain decomposition results by averaging the plurality of sets of component: