Multiple Target Localization Using Compressive Sensing

In recent years,compressive sensing has become one of the most popular topics in signal processing field.According to compressive sensing theory,if a sparse signal or a signal that features processing sparsity in a certain transform domain meets the RIP condition,it can sample signals at a rate that is much lower than the Nyquist Sample Rate and recover the original signals through certain reconstruction algorithm.Compressive sensing has solved computational complexity and memory capacity problems brought by high sample rate,which has been widely used in medical treatment,communications,military and so on.Spatial sparsity of multi-target localization makes compressive sensing theory applicable.Recently,a compressive sensing multi-target localization method based on RS SI fingerprints dictionary has been raised and achieved steady progress.In wireless sensor networks,the capacity and communications capability of wireless sensors are restricted by power,but computing center is not restricted by hardware,suggesting that capabilities of computation and communications can meet subjective requirements.The compressive sensing theory relieves the operation burden of nodes,making them lightened,but at the price that computing center needs to operate loads of signals and calculation.Since the majority of calculation of compressive sensing focuses on reconstruction algorithm,and reconstruction algorithm is operated in computing center,compressive sensing matches this condition perfectly.Given the limitations of traditional multi-target localization model based on compressive sensing,this paper proposes improved algorithm from mesh topology design,reconstruction algorithm improvement and sparse position vector operation to improve the accuracy and flexibility of location restoration.The main contributions of this paper are the following:1)Traditional rectangular mesh topology sensors cannot be deployed precisely to the mesh points,which causes errors.On account of that,I take advantage of a priori information of sensors' locations and act in contravention,proposing Delaunay mesh topology generation algorithm.With lost structural information caused by preprocessing of traditional orthogonal matching pursuit,I propose hierarchy greedy match pursuit algorithm to improve accuracy of location restoration.Diversity combining algorithm of sparse consequence vector is determined by sparse position vectors' sparsity restoration differences,which come from the comparison between the combination of HGMP algorithm and Delaunay triangular mesh topology,and the combination of HGMP algorithm and traditional rectangular mesh topology.This breaks the limitation of single algorithm.2)Build sensor platform based on gnuradio software and android device.Through applying interaction layer of software platform and hardware device,flow graph construction layer of gnuradio,communication layer of flow graph and user interface,and demonstration layer of user interface,gnuradio can be transplanted and operated on android platform,thus accomplish frequency spectrum collection and inter-device communication.