Underwater Acoustic Signal Denoising Method Based on CPo-VMD-WT

In this study, we introduce a novel denoising approach that integrates the crested porcupine optimization (CPO) algorithm with variational mode decomposition (VMD) to address the non-stationary characteristics of noisy underwater acoustic signals. The approach is refined using an enhanced wavelet thresholding (WT) method. Initially, the CPO algorithm is employed to optimally tune the parameters of the VMD optimally, facilitating the adaptive decomposition of the noisy hydroacoustic signal into multiple intrinsic modal functions (IMFs) in a non-recursive manner. Subsequently, the cosine similarity among the IMFs is computed to categorize them into signal-dominant and noise-dominant components based on a predefined threshold. The noise-dominant IMFs are then subjected to denoising through an improved WT technique. Reconstruction is performed to obtain the denoised signal. Experimental validations on both simulated signals and a public underwater dataset demonstrate the superior performance of our method compared to several existing denoising techniques, affirming its effectiveness and potential for broad application.