Underwater Noise Signal Processing Method Based on LMD Envelope Spectrum

Aiming the non-stationary characteristics of underwater noise signal, an underwater noise signal processing method based on LMD envelope spectrum is proposed in this paper, which uses the local mean decomposition (LMD) on the underwater noise signal, and product function (PF) components of some instantaneous frequencies with physical meaning are obtained. Then, the kurtosis of each PF component is calculated, and the components that contain more information of underwater target noise are selected as sensitive components according to kurtosis feature. After that, the wavelet packet decomposition and reconstruction are used for these sensitive PF components, and then the Hilbert transform is used to find the envelope spectrum of the reconstructed sensitive PF components. Through the processing and analysis of the signal collected by the experiment at the wharf, and compared with the EMD decomposition method, the experimental results show that: (1) When the underwater noise signal is processed, the result of LMD decomposition is better than that of EMD decomposition. (2) The underwater noise signal processing method based on LMD envelope spectrum can process underwater noise signal effectively, and there is an obvious spectral line in the frequency range of target sound source.