| Since sound waves travel farther than electromagnetic and light waves in complex underwater environments,underwater acoustic communication is still the most critical communication method.However,the complex sea environment and transmission channel lead to the acquisition of observation signals containing messy noise and little information,which brings difficulties to the source signal recovery in the case of a low signal-to-noise ratio(SNR).At the same time,due to the limitation of the environment and experimental cost,the number of observation signals collected by sensors is generally smaller than the number of source signals.In the above low-SNR and underdetermined conditions,it is difficult to use the traditional blind source separation method to recover the source signal accurately.Therefore,aiming to solve the problem of complex solution of mixing matrix and poor separation effect of the underwater acoustic signal,the solution in this paper is based on the "two-step method" framework of sparse component analysis theory.We propose an Underdetermined Blind Source Separation(UBSS)method based on Energy Peaks Detection(EPD)and Density Peaks Clustering(DPC).Its primary contents are as follows:(1)A wavelet threshold method based on the Optimized Gray Wolf OptimizationWavelet threshold(OGWO-Wavelet)is proposed,which realizes preprocessing of observation signals and reduces noise interference.First,given the shortcomings of the Gray Wolf Optimization algorithm(GWO),such as slow convergence speed and easy falling into local optimum,we propose the Optimized GWO(OGWO)algorithm,which improves the population initialization strategy and convergence model of the GWO algorithm.Then,the optimized OGWO is used to solve the wavelet threshold selection problem in the signal preprocessing process.The simulation results show that,compared with the GWO,the OGWO has a faster convergence speed and better convergence performance.In addition,the root mean square error of the OGWO-Wavelet is always smaller than that of the original Wavelet threshold method,which shows that the proposed method can effectively suppress the influence of noise and has a more vital signal extraction ability.(2)An UBSS algorithm based on EPD-DPC is proposed.First,for the problem that it is difficult to extract the energy peak value under low SNR,the OGWO-Wavelet algorithm is used to preprocess the observed signals,which helps extract the energy peak value.Then,use EPD to extract the time-frequency points at the energy peaks’ frequency,enhance the signal’s sparsity,and use DPC to cluster the time-frequency points to solve the mixing matrix.On this basis,aiming at the problem that the cluster centers selection of the DPC algorithm is easily affected by subjective factors,an improved DPC based on cluster center self-adaptation is proposed.Finally,the source signals are reconstructed using the L1 norm minimization method.Simulation results show that the EPD-DPC improves the mixing matrix estimation and signal separation accuracy,and the recovered source signal is closer to the real source signals.(3)Validates the effectiveness of the proposed method using both simulated and real data.First,use EPD-DPC to separate the simulated underwater acoustic test signal and underwater acoustic communication signa.The experimental results show that the source signal recovered by the proposed algorithm is least affected by noise.Then,a comparative experiment is further carried out using real data composed of whales,dolphins,and ship radiation noise.The experimental results show that the separation effect of EPD-DPC is better,and the audio data of whales and dolphins are successfully separated,the separation accuracy is significantly higher than the other two comparison algorithms.In short,the proposed method exhibits strong competitiveness in solving the problem of blind source separation of underwater acoustic signals in low SNR and underdetermined conditions.The research results have both theoretical and practical value,providing academic guidance and technical support for further improving the reliability of underwater acoustic communication. |
