Underwater acoustic signal denoising method based on DBO–VMD and singular value decomposition

With the increasingly complex marine environment, the propagation of underwater acoustic signals is often severely interfered by complex background noise, which seriously affects the signal quality and subsequently affects the detection and recognition ability of sensors for underwater targets. Regarding this challenge, this paper proposes a novel method called DBO–VMD–SVD. This method utilizes the Dung Beetle Optimization (DBO) algorithm to optimize Variational Mode Decomposition (VMD) and combines it with Singular Value Decomposition (SVD). First, the DBO method is used to optimize the VMD parameters and obtain several modal components. Second, the adaptive threshold is calculated by comprehensively considering the correlation coefficient and the multivariate permutation entropy, and it is used as the basis for selecting effective modal components and reconstructing the signal. Finally, the obtained signal is further processed using the SVD method. The innovation of the method proposed in this article is mainly reflected in (1) using the DBO algorithm to avoid manually setting VMD parameters based on experience, (2) using adaptive threshold to select intrinsic mode function (IMF) components, and (3) combining the VMD method with the SVD method. In an effort to verify the performance of the DBO–VMD–SVD method, it was used to denoise Lorenz signals at different signal-to-noise ratios (SNRs) and compared with methods such as VMD and VMD–SVD. The results show that our proposed method has a more significant denoising effect, exhibiting certain advantages in evaluation indicators, such as SNR and root mean square error. After processing with this method, the geometric structure of the signal becomes more regular. Moreover, it was applied to the noise reduction of real underwater acoustic signals. The simulation results show that after processing by this method, noise intensity, correlation dimension, and permutation entropy all decreased. Taking signal-1 as an example, the three evaluation indicators decreased by 14.5%, 14.7%, and 41.1%, respectively. In addition, the phase diagram became more regular, more clearly restoring the topological structure of the chaotic attractor. This paper provides a novel approach for improving underwater target detection and recognition capabilities, optimizing underwater acoustic communication quality, and other related research.