无线传感器网络下基于压缩感知的多目标分层贪婪匹配定位

针对基于压缩感知（Compressive sensing，CS）的多目标定位问题，通过分析多目标场景中的隐含结构信息，本文提出一种层级的贪婪匹配追踪定位算法.该算法首先获得多目标在网格化空间中的可能位置作为全局估计层，然后利用该全局估计信息作为稀疏恢复层的输入信息，在网格化空间中重构多目标位置矢量.本文证明了文献中广泛采用的基于正交化的预处理方式实质上降低了信噪比（Signal to noise ratio，SNR），从而降低了定位性能.本文通过全局估计，预先排除了不可能的位置，等效于从观测子空间中分离出信号子空间，从而降低了观测噪声的影响.通过理论分析与计算机仿真，表明所提算法具有线性复杂度且在相同信噪比下具有更高的定位正确率和定位精度.

Hierarchical Greedy Matching Pursuit for Multi-target Localization in Wireless Sensor Networks Using Compressive Sensing

This paper addresses the problem of multi-target localization in wireless sensor networks using compressive sensing (CS). We first analyze the ample implicit structured information contained in the multi-target localization scenario. Then, a hierarchical greedy matching pursuit (HGMP) approach for multi-target localization is proposed. In the proposed HGMP algorithm, the possible positions of targets in the meshing space are obtained as the global estimation layer, and subsequently the global estimation information is used as the input information to the sparse recovery layer to reconstruct the multi-target localization vector in the meshing space. Moreover, we prove that the orthogonality-based preprocessing operation widely adopted in the literature reduces the signal-to-noise ratio (SNR), degrades the localization performance, a problem that has never been addressed before. Through the global estimation layer, impossible positions are preliminarily removed, which is equivalent to separating the signal subspace from the observation subspace, thus reducing the influence of the observed noise. Finally, theoretical analysis and computer simulations show that the proposed algorithm enjoys a linear computational complexity and a higher localization accuracy at the same SNR.