A new denoising method based on decomposition mixing of hydro-acoustic signal

Abstract

Hydro-acoustic signal (HAS) contains abundant information. HAS denoising is of great significance for accurate identification of underwater targets, research of military detection equipment and monitoring of ocean ecological environment, and has good civil value and military application value. Because of complex characteristic of ocean circumstance and the temporal change characteristic of hydro-acoustic channel, HAS denoising faces great challenges. To denoise HAS effectively, a new denoising method based on improved variational mode decomposition by jumping spider optimization algorithm (JVMD), weighted fractional order slope entropy (WFrSlEn), dynamic adjacent threshold of correlation coefficient (DATC), improved one dimension (1-D) Gaussian-Laplacian smoothing (IGLS) and adaptive interval correlation wavelet transform (AICWT), named JVMD-WFrSlEn-DATC-IGLS-AICWT, is proposed. JVMD is proposed for the problem that the number of decomposition layer K and penalty factor α cannot be selected adaptively in variational mode decomposition (VMD). WFrSlEn is proposed for the problem that decomposition components are difficult to divide precisely. DATC is proposed for the problem that threshold of correlation coefficient (CC) is selected. IGLS is proposed to preserve more signal details. AICWT is proposed to solve the problem that wavelet transform cannot automatically select appropriate parameters. Firstly, the original signal is decomposed to many intrinsic mode functions (IMFs) with JVMD. Secondly, WFrSlEn and DATC distinguish pure IMFs, mixed IMFs and noisy IMFs. IGLS is performed on mixed IMFs and AICWT is performed on noisy IMFs. Finally, two processed IMFs are reconstructed with pure IMFs to obtain final denoised signal. The results show that the proposed denoising method can improve signal-to-noise ratio by 8∼10 dB for chaotic signal, make phase diagram clearer and smoother, and can effectively suppress noise for HAS.