Research On Autonomous Detection And Identification Method Of Underwater Target Radiated Noise Line Spectrum

The low-frequency line spectrum of radiated noise is an important feature of underwater targets.The passive sonar of manned platforms such as ships usually uses large aperture array to detect it,and completes the extraction and identification of weak target line spectrum with manual assistance.However,the size of the underwater unmanned platform is small,and the spatial processing gain and resolution of the underwater acoustic array are limited.The conventional methods usually cannot meet the requirements of the traditional weak and low frequency line spectrum detection and identification processing.In addition,the remote interaction of information of underwater unmanned platform is difficult.The array metadata cannot be transmitted back to the ship or shore based platform,and can only transmit limited information such as line spectrum detection results,which requires the passive sonar of underwater unmanned platform to have strong autonomous processing ability.According to the requirements of passive sonar target detection of underwater unmanned platform,this paper studies the processing and implementation method of autonomous detection and identification of underwater target radiated noise line spectrum.The main research contents are as follows:Aiming at the problem of weak line spectrum detection under strong nonwhite background noise,an LMK(least mean kurtosis)adaptive line spectrum enhancement method based on lp norm is proposed.This method uses lp norm instead of l1 norm commonly used in sparse driven adaptive enhancement method as constraint term,and uses the negative kurtosis of error signal as cost function to suppress Gaussian colored noise component.The simulation results show that the performance of adaptive line spectrum enhancement method under color noise background can be improved by introducing negative kurtosis into the cost function.When p is less than 1,the performance of lp norm driven adaptive enhancement algorithm is better than that of l1 norm driven enhancement algorithm;When the input signal-to-noise ratio is-20 d B,the signal-to-noise ratio gain of l1/2--LMK method is0.6d B and 3d B higher than that of l1/2-ALE method in white noise background and color noise background respectively.Aiming at the problem that the conventional sparse processing line spectrum enhancement algorithm needs to manually specify the prior information of sparsity,an autonomous line spectrum extraction method based on SAMP(Sparse Adaptive Matching Pursuit)algorithm is proposed.Firstly,the sparse model of line spectrum is solved by SAMP algorithm.The real frequency domain sparsity of line spectrum signal is approximated iteratively and the preliminary line spectrum signal is obtained.At the same time,the line spectrum signal leakage problem caused by broadband noise spectrum peak is improved by using the method based on sub band decomposition and background equalization;Secondly,an adaptive threshold determination criterion is designed to identify the "pseudo spectral peak" caused by the random noise in the solution results of sparse model.The simulation results show that under the same time length,this method can extract two line spectra that cannot be distinguished by the traditional periodic graph method.At the same time,compared with the periodic graph method,the line spectrum detection method based on SAMP algorithm extracts fewer "pseudo spectrum peaks" and lower false alarm rate.Aiming at the problems of low spatial gain and weak azimuth resolution of conventional beamforming of small-scale array,a line spectrum identification method based on virtual array elements is designed.This method obtains high-precision signal direction results by expanding the number of virtual array elements,and suppresses the interference signal by inverse beamforming.The simulation results show that this method can improve the spatial gain and resolution performance of small-scale array to a certain extent,suppress the interference signal and extract a relatively pure target line spectrum.Based on the GPU computing platform,an autonomous line spectrum extraction and processing system is designed and implemented.Using the method proposed in this paper,the real-time array processing of radiated noise signal and the autonomous detection and identification of low-frequency line spectrum can be completed.The results of sea test data processing show that the line spectrum autonomous extraction system can successfully realize the autonomous detection of 6 line spectrum signals emitted by the target sound source,and the number of error line spectra detected is less than that of the periodic graph method under the same conditions.