A distributed sensor array network is studied, where sub-arrays are placed on those distributed observation platforms. In this model, bearing-only source localization is characterized in terms of direction of arrival (DOA) if the sources are far from the entire network, while their locations in the predefined Cartesian coordinate system can be obtained for the near-field case. For wideband signals, the focusing algorithm is applied at each sub-array to form an equivalent single frequency signal model. Then, a compressive sensing (CS) based DOA estimation method employing the group sparsity concept is proposed for far-field sources with the information acquired by all the platforms processed as a whole. This concept is further extended to near field, and a group sparsity based method to localize the near-field sources is derived. The proposed solutions are applicable for both uncorrelated and coherent signals, and the corresponding Cramér-Rao Bounds (CRBs) are derived. Compared with the maximum likelihood estimator (MLE) of forming the final result through a fusion process, where separately estimated unreliable bearing result at even one observation platform would spoil the overall performance, improved performance is achieved by both proposed methods. It is noted that only the covariance matrix in lieu of data samples at each platform is required for centralized processing, and therefore the increase of the data exchange workload among platforms is rather limited.

中文翻译：

---

基于分布式传感器阵列网络的远近场源群稀疏定位

研究了分布式传感器阵列网络，其中子阵列放置在这些分布式观测平台上。在该模型中，如果源远离整个网络，则仅轴承源定位的特征在于到达方向 (DOA)，而对于近场情况，可以获得它们在预定义笛卡尔坐标系中的位置。对于宽带信号，聚焦算法应用于每个子阵列，形成等效的单频信号模型。然后，针对远场源提出了一种基于压缩感知（CS）的DOA估计方法，该方法采用群稀疏概念，并将所有平台获取的信息作为一个整体进行处理。这个概念进一步扩展到近场，并导出了一种基于群稀疏性的近场源定位方法。所提出的解决方案适用于不相关和相干信号，并推导出相应的克拉梅-拉奥边界 (CRB)。与通过融合过程形成最终结果的最大似然估计器 (MLE) 相比，即使在一个观测平台上单独估计不可靠的方位结果也会破坏整体性能，这两种方法都提高了性能。需要注意的是，集中处理只需要协方差矩阵代替各个平台的数据样本，因此平台间数据交换工作量的增加是比较有限的。与通过融合过程形成最终结果的最大似然估计器（MLE）相比，即使在一个观测平台上单独估计不可靠的方位结果也会破坏整体性能，这两种方法都实现了性能的提高。需要注意的是，集中处理只需要协方差矩阵代替各个平台的数据样本，因此平台间数据交换工作量的增加是比较有限的。与通过融合过程形成最终结果的最大似然估计器 (MLE) 相比，即使在一个观测平台上单独估计不可靠的方位结果也会破坏整体性能，这两种方法都提高了性能。需要注意的是，集中处理只需要协方差矩阵代替各个平台的数据样本，因此平台间数据交换工作量的增加是比较有限的。