Multiple Target Localization Algorithm Research Based On Compressive Sensing In Wireless Sensor Network

With the development of the wireless communication, digital electronic technology and micro-electromechanical technology, wireless sensor network (WSN) has been rapid development and extensive application. Localization is the most essential problem since the location information is the important premise for many applications and services. As WSN is a kind of low-power and low-cost network, localization technology must have the advantages of low-cost, low-power, low-complexity and high-accuracy. However, the existing localization algorithms have limitations on adaptability. Therefore, it has theoretical significance and applicable value to improve the adaptability of localization algorithms.This paper mainly has studied multi-targeting localization algorithm in wireless sensor network. To solve localization problems under the different network conditions, the paper offers a few methods for these problems in wireless sensor network by using compressive sensing, which is the frontier theory in signal processing field. Specifically, the paper's main research results include the following aspects:Firstly, An Improved Greedy Matching Pursuit Algorithm for Multiple Target LocalizationConventional distance based target location methods position the targets at the centers of certain grids which represent partitions of the target area. This will lead to degraded location accuracy or increasing localization errors when the targets locate around the grid borders. By exploiting the multi-resolution analysis (MRA). we propose modified GMP based multi-targets location method. The proposed method provides better solution in target location by infinitely approaching to the real positions of targets. Therefore it achieves promising location accuracy with decreased location error.Secondly, Sparse target localization via StOMP in wireless sensor network.In the existing target localization algorithms based on compressive sensing. the measurement matrixes dont satisfy Restricted Isometry Property (RIP). To solve this problem.a novel localization algorithm-sparse target localization via StOMP is proposed. The proposed algorithm formulates the target localization problem with compressive sensing model via grid process. More specific, QR decomposition is first used to get a new measurement matrix which satisfies RIP property. Then StOMP is exploited to conduct sparse representation. The proposed method is testified on a simulated network. The experimental results show that our proposed method shows promising performance in terms of both target localization accuracy and processing speed, compared with sparse target localization via orth preprocessing.