Research On Key Technologies Of Frequency Domain Analysis Of Passive Sonar Signal For Moving Target

Detection and estimation of line spectrum in underwater target radiated noise is an important research area of passive sonar signal processing.In general,detection and parameter estimation of line spectrum by passive sonar are performed under strong background noise and strong interference.It is necessary to increase the integration time to improve frequency resolution and obtain higher signal processing gain.However,the frequency drift of line spectrum and target motion limit the time length of integration processing and signal processing gain provided by traditional passive sonar signal processing methods.In this paper,we have conducted in-depth research on how to improve the spectral resolution of power spectrum analysis,suppress the side lobe and reduce the cross-correlation loss caused by target motion with limited data sample length.First,for the problem of power spectrum analysis of moving targets under strong line interference,a high-resolution power spectrum estimation method based on deconvolution algorithm in frequency domain is proposed.According to the theory of classical spectrum estimation,the power spectrum of a finite-length sample can be obtained by convolution calculation of the power spectra of an infinite-length sample and a window function.It is proposed in this paper that the power spectrum of an infinite-length sample can be obtained by taking the power spectrum of a finite-length sample into deconvolution with the power spectrum of the window function.The deconvolved power spectrum has higher frequency resolution and lower side lobes.The proposed method can improve the frequency resolution,reduce the spectrum leakage to suppress the side lobe,and reduce the masking of strong line spectrum to the adjacent weak target line spectrum.The number of iterations of the deconvolution algorithm used in this paper is the main factor that affects the frequency resolution and sidelobe suppression performance for spectral estimation.In simulation and experimental data processing,the performance of MVDR(minimum variance distortionless response),CS(compressed sensing)and the proposed power spectrum estimation method based on L-R deconvolution algorithm are compared in this paper.It can be seen in the processing results that the power spectrum estimation method based on deconvolution can effectively improve the frequency resolution and reduce the influence of side lobes.It is suitable to be applied for spectral estimation of passive sonar signals under strong line interference and time-frequency analysis using finite length signals radiated by moving targets.Secondly,a cross-power spectrum model of target radiated signals received by two hydrophones is established,and a two-step method is proposed to estimate the Doppler coefficient(scale factor)and the time delay of the cross-power spectrum obtained by the received signals of two hydrophones.The Doppler coefficient is first estimated using the amplitude spectra of the received signals,the estimated Doppler coefficient can be used to compensate the received signal,and then the phase of the cross-power spectrum obtained by compensated received signals can be used to estimate the time delay.The algorithm converts the parameter estimation from two-dimensional search of Doppler coefficient and time delay to two one-dimensional searches along the Doppler axis and the time delay axis,which not only effectively reduces the calculation amount of parameter optimization,but also realizes the decoupling of Doppler coefficient and time delay parameter estimation to avoid the mutual interference in the parameter estimation process and improve the accuracy of the parameter estimation results.The results show that the two-step method for Doppler coefficient and time delay estimation based on cross-power spectrum can effectively improve the accuracy of the time delay estimation,thereby improving the performance of cross-power spectrum for direction finding.Finally,a segmented data model is established about Doppler coefficient and time delay of the signals received by two hydrophones with target motion.Due to target motion,the crosspower spectrum of the target radiated acoustic signal in each pair of data segments has frequency drift and phase changing.It is difficult for these cross-power spectra to be used for coherent accumulation.In order to solve this problem,a coherent integration method of crosspower spectrum with Doppler and time delay compensations is proposed in this paper.With this method,the Doppler coefficient and time delay of cross-power spectrum of different data segments can be estimated and then appied for compensations.The compensated cross-power spectrum of each data segments can be used for coherent integration.Performance of the propsed coherent integration method of compensated cross-power spectrum,MUSIC(multiple signal classification),coherent accumulation and incoherent accumulation algorithms of cross power spectrum is compared in simulation and experimental data processing and analysis.It is shown that the compensated cross-power spectrum coherent integration method can overcome the influence of Doppler effect,reduce the cross-correlation loss,increase the time length of integration processing for high processing gain,and then improve the detection performance of weak line spectrum of moving targets.