Font Size: a A A

Research On Direction-of-arrival Estimation Algorithms Based On Sparse Representation Of Signals

Posted on:2017-02-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:X Y LuoFull Text:PDF
GTID:1108330485988402Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
The direction-of-arrival(DOA) estimation for spatial signals are the important basis for localization, navigation, interference, imaging and so on. Compared with the existing algorithms of DOA estimation, the subspace-based algorithms are more efficient. However, there is a strong restriction for the rank of received data matrix or sampling covariance matrix in this kind of algorithms. When the rank is smaller than the number of signals, the spatial smoothing techniques are needed to revise the rank. As a result, the utilization ratio of sensors in this kind of algorithms is low, which leads the poor performance of DOA estimation. With the fast development of the theory of compressed sensing, the theory of sparse representation of signals is exploited in DOA estimation. Compared with using the subspace-based algorithms, the impact on DOA estimation caused by the rank of received data matrix or sampling covariance matrix can be reduced by using the sparse representation-based DOA estimation algorithms. Thus, a deep research of the sparse representation-based DOA estimation algorithms is carried out in this dissertation. The specific research contents and contributions can be summarized as the following four aspects.1. The research of the sparse reconstruction algorithms in the sparse representation-based DOA estimationThe method exploiting 1l norm to replace 0l norm is widely used in sparse reconstruction among the existing sparse representation-based DOA estimation algorithms, while it is not robust due to the difference in the definitions of 1l and 0l norms. The performance of this method is poor especially when the sparsity of signal model is reduced by the noise and the insufficient statistics. To solve this problem, an improved 1l norm-based algorithm is proposed in the chapter 3 of this dissertation. In this algorithm, the reweighted 1l norm is used to replace the conventional 1l norm to achieve the sparse reconstruction. According to the theoretical derivation, it can be find that the difference in the definitions of 1l and 0l norms is reduced by this algorithm effectively, which improves the accuracy of DOA estimation. The performance of the proposed algorithm is remarkably better than those of the conventional 1l norm-based methods especially in the cases of low signal-to-noise ratio and small number of snapshots.2. The researches of revising the model of sparse representation and the sparse reconstruction algorithms in the off-grid caseUsing grid processing to obtain the overcomplete basis matrix is the core idea in sparse representation-based DOA estimation algorithms. However, grid processing may lead to the off-grid problem. When all true angles do not locate on the grid points of the overcomplete basis matrix, most of the existing algorithms are invalid. To solve this problem, in the chapter 4 of this dissertation, the first-order Taylor series is used to revise the conventional sparse representation model of array covariance matrix, and then an alternating iterative algorithm is proposed to achieve the sparse reconstruction under this revised model. This alternating iterative algorithm is constructed between a minimization of 1l norm and a least square solution. Compared with conventional methods, the accuracy of proposed algorithm for off-grid DOA estimation is higher.3. The research of the DOA estimation algorithm in multipath propagationsThe DOA estimation may be under the coexistence of both uncorrelated signals and different groups of coherent signals due to the multipath propagations. The utilization ratio of sensors is reduced in the case that the uncorrelated signals and different groups of coherent signals cannot be distinguished. To solve this problem, two algorithms are proposed in the chapter 5 of this dissertation. Firstly, the eigenvalue method is proposed by exploiting the properties of eigenvalues. Secondly, the disadvantage of eigenvalue method is analyzed, and an improved method named eigenvector method is proposed by using the correlation of eigenvectors. Compared with the eigenvalue method, the ability of eigenvector method to distinguish the uncorrelated signals and different groups of coherent signals is better in the case of small number of sensors. Finally, on the basis of distinguishing the uncorrelated signals and different groups of coherent signals, the sparse representation of each group of coherent signals is achieved, and the sparse reconstruction is then accomplished. Compared with the conventional methods which combine subspace decomposition and spatial smoothing techniques, the proposed methods improve the utilization ratio of sensors and outperform the conventional methods in multipath propagations.4. The research of the two-dimensional(2-D) DOA estimation algorithm based on L-shaped arrayStudying the algorithms for 2-D DOA estimation based on L-shaped array is one of the most popular research points in DOA estimation. For the subspace-based methods,reducing the computational burden is the main purpose. Additionally, the existing methods are lack of the research of using sparse representation of signals to achieve 2-D DOA estimation. To solve these two problems, firstly, a subspace-based 2-D DOA estimation method with low complexity is proposed in the chapter 6 of this dissertation, and the left and right singular matrices are simultaneously exploited to achieve 2-D DOA estimation without pair matching. Compared with conventional methods, the complexity of this method is reduced without loss of the accuracy of DOA estimation. Secondly, the sparse representation-based method is extended to 2-D DOA estimation, and the joint sparse representation of cross-correlation matrix for elevation and azimuth angles is achieved by using L-shaped array. Based on this model, the conventional orthogonal matching pursuit method is extended to 2-D space to achieve the joint estimation of elevation and azimuth angles. Finally, simulation results show the efficient of proposed methods.
Keywords/Search Tags:direction-of-arrival estimation, sparse representation of signals, off grid, multipath propagations, cross-correlation matrix
PDF Full Text Request
Related items