中文摘要：

针对已有的基于特征矢量稀疏重构的DOA估计方法中需要选取平衡残差项与结果稀疏性的正则化参数的问题，提出了一种新的特征矢量稀疏重构求解方法。首先在有限快拍数条件下，由特征矢量估计误差的统计分布特性，得到残差项大置信度的置信区间；再以此置信区间作为约束条件，以解矢量的11范数作为最小化的目标函数，由此利用二阶锥规化求解时避免了正则化参数的选取。理论分析与仿真实验表明本文算法计算复杂度低；能够对非相干及相干信号的DOA进行估计，且具备很好的解相干性能；低信噪比条件下，对DOA的估计误差随着采样快拍数的增大而减小。

英文摘要：

In the existing DOA estimation approach based on sparse reconstruction of eigenvector, it is necessary to select regular parameter which balances residual with sparsity of solution. In order to avoid selecting regular parameter, a novel approach to solve the sparse reconstruction of eigenvector is proposed. Firstly, in the case of finite snapshots, the large confidence interval of residual is got by the statistical distribution of eigenvector estimation error. Secondly, residual is constrained in the confidence interval and 1 norm of solution is taken as target function for minimization. Therefore, the selection of regular parameter is unnecessary when using the second order cone to solve it. Theoretical analysis and experimental results show that the approach has low computation complexity. It can estimate DOA of incoherent and coherent signals and the performance of exacting coherent signals is excellent. In the case of low SNR, DOA estimation error decreases along with increase of snapshots.