Research On High Resolution DOA Estimation Algorithm For Wideband Array Signal

The DOA estimation is the key content of array signal processing and has important applications in the military and national economy.As the signal environment becomes more and more complex,the application of wideband signals is becoming more extensive.Therefore,it is of great significance to study the high resolution DOA estimation algorithm of wideband signals.In this paper,we focus on the current problems of DOA estimation algorithm for wideband signals.The main work and innovation points include the following aspects:1.Aiming at the problem that the performance of the Gerschgorin disks method was degraded when the adjustment factor was set improperly in source number estimation,this paper proposed a modified Gerschgorin disks method based on total least squares fitting.This method kept the radius of the disk as the fitting points of straight line so that the fitting discrepancy of the signal disk radius was much larger than that of the noise disk radius,and a ratio-step criterion was developed to estimate the source number so that no artificial adjustment factor was set.The performance of the TLS-GDE algorithm was verified by simulation.The results show that the robustness of the TLS-GDE is improved and the performance of the TLS-GDE is better than that of the GDE in low SNR.2.In order to solve the problem of incoherent signal subspace method under nonuniform noise condition,two kinds of denoising methods,such as Reduced Covariance Matrix(RCM)and Rank and Trace Minimization(RTM),were proposed.The RCM algorithm eliminated the noise component by defining an attenuation covariance matrix to achieve the effect of denoising.The RTM algorithm was based on the method of minimizing the rank of the constraint matrix and was worked out by the convex optimization toolbox after minimizing the kernel norm.Finally,the simulation results show that the two algorithms can effectively suppress the nonuniform noise.3.The test of orthogonality of projected subspaces(TOPS)utilized the orthogonality of subspace of multiple frequency points of wideband signal to complete the directions of arrival(DOA)estimation.The performance of TOPS was poor at low signal-to-noise ratio,and depended on the choice of reference frequency point.Besides the TOPS often exhibited pseudo-peak.This paper proposed a modification(FTOPS)based on the idea of focus to overcome the shortcomings of TOPS.Firstly,the method of reduced covariance matrix(RCM)was utilized to eliminate the noise.And then the signal subspace at each frequency point was focused on the signal subspace of any reference frequency.Finally,the DOA estimation was completed using the orthogonality between the signal subspace of the reference frequency and the orthogonal projection matrix of the array direction vector.The simulation results show that the FTOPS does not depend on the selection of reference frequency,and can effectively eliminate the pseudo peak,and the performance of FTOPS is superior to the traditional TOPS algorithm at low SNR.4.The maximum likelihood method was a nonlinear optimization problem in the estimation of signal direction of arrival,and there was a problem that the computational complexity was large and it was easy to fall into the local optimal solution.At the same time,it needed a certain number of samples to guarantee the performance.In view of this problem,this paper presented an improved algorithm.First,this method discretized the airspace and obtained the sparse model under the sparse sampling condition.After the convex relaxation,the convex parameter was obtained by using the convex heuristic algorithm.Then the sparse solution was taken as the initial point.At this point,the first-order Taylor expansion was performed on the array direction matrix,which was transformed into a sequence quadratic programming problem.And then the gradient descent method was used to alternately optimize the direction parameter vector and signal parameter vector.The performance of the proposed method was verified by simulation.The results show that the algorithm can converge to the global optimal solution quickly,and it has high estimation precision.