Research On Spatial Shift Variant Deconvolved Beamforming

As an important means to improve the performance of sonar target detection,high-resolution beamforming technology has been a research hotspot of scholars at home and abroad for decades.However,due to the large array manifold error and the difficulty to guarantee the signal stability for a long time,resulting in the practical application performance of many traditional high-resolution array signal processing methods.In recent years,high-resolution array processing methods with high robustness and real-time(small snapshot,short sample)characteristics have become a new research hotspot in array signal processing.The beam power output of conventional beamforming can be expressed as the convolution of the array beam pattern and the source distribution function.The array beam pattern is limited by the aperture,resulting in limited resolution.According to the duality of time sampling and spatial sampling,time series deconvolution can be applied to spatial array.The use of deconvolution technology to remove the influence of the directivity of the array and get rid of the limitation of the physical aperture provides a new idea for the robust and high-resolution signal processing of the underwater array.Deconvolved beamforming(d Cv)is a robust high-resolution beamforming method which yields narrow beam and low sidelobe.Its applications in uniform line array and circular array have been effectively verified,showing robust high-resolution performance.However,the traditional d Cv algorithm which shows good performance in low SNR is not suitable for the array with shift-variant beam pattern.In this paper,aiming at the robust and high-resolution processing requirements of underwater acoustic array signal processing of vector array and other spatial shift model array,the robust and high-resolution d Cv algorithms suit for shift-variant array are studied.First,a variety of array convolution/deconvolution models are established,and on this basis,a variety of deconvolution beamforming algorithms are studied,focusing on the solution of deconvolution beamforming under the condition of spatial shift-variant model.The research content of this paper mainly includes:The research on variety of array convolution/deconvolution theory modeling and convolution model classification.According to the spatial shift-variance characteristics of underwater acoustic array,the convolution models of various commonly used underwater arrays are derived and classified.This paper introduces the convolution/deconvolution model in the time domain system,then makes a time-space analogy,and introduces the problem of d Cv in array processing.This paper analyzes and deduces the convolution models of various commonly used underwater array forms,which is based on the beam pattern(which can also be called The Point Spread Function,PSF)of the array is classified and summarized.The types of convolution models are classified and divided into one-dimensional shift-invariant deconvolution problems,one-dimensional shift-variant deconvolution beamforming problems,two-dimensional(2D)shift-invariant deconvolution beamforming problem and 2D shift-variant deconvolution beamforming problem.This thesis studies three typical one and two-dimensional spatial shift-variant d Cv problems,which are vector sensor line array,2D near field beamforming of sound pressure array and 2D spatial-temporal problem of circular arc array.The research on deconvolved beamforming of vector array based on one-dimensional spatial shift-variant model.It includes the theoretical shift variant convolution model of vector array and the research of one-dimensional shift variant deconvolution algorithms.Aiming at the problem that the beam pattern of vector sensor line array is shift-variant,a variety of d Cv algorithms of one-dimensional shift-variant model are studied,an extended RL algorithm(Extended Richardson Lucy,Ex-RL)is proposed to improve the inaccurate azimuth near the endfire direction and peak to sidelobe level.It is compared with two existing deconvolution algorithms suitable for shift variant model,Nonnegative Least-squares(NNLS)and DAMAS(Deconvolution Approach for the Mapping of Acoustic Sources).The simulation analysis compared the main lobe width,peak to side lobe ratio,array gain and robustness of various algorithms under different iteration times,different signal-to-noise(SNR)ratios,and different environmental noise fields.Simulation and experimental data prove that the Ex-RL algorithm has the best overall performance under the low SNR condition.The research on deconvolved beamforming based on 2D spatial shift variant model for near-field acoustic image measurement.It includes the near-field 2D spatial shift variant convolution/deconvolution theoretical model and the deconvolution algorithms.Aiming at the problem of low sound source localization accuracy at low frequency,a 2D extended RL algorithm is proposed to solve the two-dimensional shift model of near-field acoustic images.First,the near-field 2D convolution model of acoustic image measurement is introduced.The deconvolution process of solving the sound source distribution function is a typical 2D shift-variant deconvolution problem.And use d Cv technology to improve the positioning accuracy of the source position.On this basis,a 2D Ex-RL deconvolution beamforming algorithm is studied,which is suitable for solving the 2D shift variant model of the near-field acoustic image.The performance of four acoustic image measurement algorithms,CBF,Ex-RL,DAMAS,and NNLS,are compared and analyzed through simulation and experimental data.The results show that the proposed Ex-RL algorithm is comparable to the existing DAMAS and NNLS under the condition of low SNR.It has the advantages of high positioning accuracy and narrow focus peak size.The research on spatiotemporal 2D deconvolution beam imaging technology for active acoustic imaging.It includes the active sonar spatiotemporal 2D convolution/deconvolution theoretical model and the deconvolution algorithms.Aiming at the problem of low range resolution of active sonar,a 2D deconvolution beam imaging method based on the two physical dimensions of range and azimuth is proposed.In the azimuth dimension,the conventional beamforming spatial spectrum output can be expressed as the convolution result of the array beam pattern and the position information of the point target.Therefore,the azimuth information can be obtained by deconvolution,which can effectively improve the accuracy of the target azimuth estimation.In the distance dimension,the cross-correlation between the received signal and the transmitted signal of the array is the convolution of the autocorrelation of the transmitted signal and the impulse response function of the target reflection channel.The d Cv can be applied in the time domain to estimate the time delay/distance information of the target reflection with high accuracy.The near-field planar array beamforming is performed for each distance to obtain the range and azimuth two-dimensional acoustic image of the conventionally processed active imaging sonar.Then,on this basis,the original image is subjected to deconvolution processing in two different dimensions of orientation and distance to obtain high-resolution measurement results.