Study On Target Direction Finding Methods In MIMO Radar Based On Covariance Matrix Sparse Representation

At the beginning of the 21st century,as a new type of radar,multiple-input multiple-output?MIMO?radar attracted much attention of scholars for its superior performance compared to traditional radars.Target direction finding problem is one of the important contents of MIMO radar.At present,a series of methods have been proposed for MIMO radar target direction findings,such as multiple signal classification?MUSIC?methods and the estimating signal parameters via rotational invariance techniques?ESPRIT?methods,etc.However,the traditional direction finding methods are limited by the subspace theory framework and its performance cannot be further improved.In recent years,the sparse signal representation theory has brought new directions and ideas to the research of target direction findings.The use of sparse signal reconstruction to obtain the target's angle of arrival can reduce the direction finding error,and solve the problem of reduced accuracy and resolution of traditional subspace methods in the case of single-snapshot?SMV?and coherent targets.Firstly,the classical sparse representation method,i.e.,singular value decomposition method based on l₁ norm?l₁-SVD?,is analyzed and compared with the traditional method via simulation.Based on it,the covariance matrix sparse representation method based on l₁-norm?l₁-SRACV?is further studied in this thesis aiming at the shortcomings of the l₁-SVD method such as the need for known prior information,e.g.,the number of targets.The l₁-SRACV method can improve the accuracy of the target direction findings by processing the target covariance,and can achieve the target direction findings without knowing the prior information.However,there are still many problems in this method:?1?When the sparse matrix is reconstructed in the case of multiple snapshots,it will lead to high complexity,large computation quantity,long calculation time and so on.?2?The real point may not fall on the grid point in practical applications,which will lead to the base mismatch problems.Aiming at the above problems,covariance matrix theory and sparse reconstruction method are used as tools in this thesis to conduct intensive research on MIMO radar target direction finding methods.Since uniform circular arrays can achieve 360-degree ambiguity-free direction finding,this thesis focuses on the MIMO radar target direction finding methods based on the sparse representation of covariance matrix under uniform circular array.The innovative research works of this thesis are as follows:In terms of the problem of large computation amount in multi-snapshot situation,a low complexity direction finding method is presented for MIMO radar based on covariance matrix sparse representation.After the multi-measurement vector?MMV?sparse representation model of array covariance matrix is obtained by SRACV method,the vectorization is used to transform the MMV model into a single measurement vector?SMV?model.Then,the target direction finding is achieved by using the decomposition transformation of Khatri-Rao product matrix and searching the optimal solution through convex optimization toolbox.Finally,the simulation results show the superiority of this method.With respect to the problem of base mismatch in traditional sparse representation method,a grid-off target direction finding method for MIMO radar based on sparse representation of covariance matrix is proposed in this paper.The virtual grid-off procedure is realized by Bayesian interpolation.Firstly,a steering vector with off-the-grid parameters is proposed to construct a grid-off target direction finding model based on covariance matrix.Then Bayesian inference is performed by covariance signal,noise and off-the-grid parameters,and the joint probability density function is obtained.Finally,the hyperparameters are solved iteratively by applying the expectation maximization algorithm,and the target direction finding is achieved by peak search.The simulation results show that the method is suitable for both single snapshot and multi-snapshot,and reduces the error caused by base mismatch of traditional sparse representation methods.