Design Of Compressive Sensing Measurement Matrix And CS Application In Wireless Sensor Networks

The wireless sensor network(WSN),with the characteristics of large-scale and strong self-organization,is widely used in environmental monitoring in military,agricultural,industrial and other fields.However,the energy limited characteristic of its sensor nodes limits the large-scale and long-term deployment of WSN in practical applications.With the expansion of WSN scale,more and more data are lost,which has become the key problem and challenge of accurate reconfiguration.Therefore,it is an urgent and important problem to design an effective algorithm to ensure the accurate reconstruction of WSN data collection with low energy consumption.It is found that Compressive Sensing(CS)is a powerful and general technology to estimate missing data,and its sampling characteristic can just improve the problems of excessive data transmission,high redundancy and limited node energy in WSN.As a new theoretical framework of information acquisition and processing,compressive sensing can recover signals from less linear and non-adaptive measurements by reducing the dimension of the signal based on the sparsity and compressibility of the signal.In particular,we find that in order to realize the "lossless" information perception and recovery of CS theory,we not only need to describe the signal sparsity more accurately,but also need a good performance measurement matrix.The main contents of this paper are as follows:(1)A method of measurement matrix construction with good structure and meeting Welch bound is proposed.The design of measurement matrix should meet RIP constraint,which is a problem of combination complexity.Therefore,it is our research focus to simplify the construction process and design the measurement matrix with good structure and good performance.This paper presents a new construction method,which is mainly as follows:First,simplify the verification of RIP constraint,start from Spark property,extend to matrix correlation,and find a suitable lower bound Welch bound for it.This method is used as the starting point to construct,which provides the lower bound of correlation for the uniqueness of dimension reduction of sparse signals,and strictly guarantees the uniqueness of measurement vector.Secondly,the matrix constructed by this method has good structure by combining with Kronecker product.Meanwhile,in the process of using Welch bound screening,the randomness of matrix is reduced to some extent.To prove the feasibility of the method,the proposed method is first concreted,the matrix HWKM is constructed and its properties are verified.Secondly,the matrix HWKM is compared with the traditional Gaussian random matrix,partial Hadamard matrix and the chaos matrix which has been excellent in recent years.The experimental results show that the reconstruction error,SNR and time complexity of the matrix constructed by this method are better than the above matrix in one-dimensional signal and two-dimensional image restoration.(2)Based on the measurement matrix construction method designed in this paper,propose a hierarchical clustering data collection algorithm with tendency based on cluster structure.The existing WSN data collection algorithms have problems such as low data compression rate,high energy consumption,unbalanced load and high time complexity.Therefore,this paper makes the following improvements based on the clustering data collection algorithm based on CS theory: First,analyzing the energy of network model is,aiming at the problem that traditional algorithm can't adaptively select the optimal cluster number and the existing "hot area",propose a clustering scheme based on hierarchical clustering.Secondly,a tendency check cluster head is proposed to solve the problem of unbalanced energy consumption of nodes,and entropy weight method is used to select the functional node selection scheme of sampling nodes.Meanwhile,considering the influence of routing on network energy consumption and stability,combined with "hot zone",propose the scheme of changing relay nodes in different stages to further balance cluster head energy consumption.Finally,based on the data recovery framework with CS and PCA,the matrix is improved to HWKM constructed by the method proposed in this paper.The Kronecker product structure can be effectively combined with the time and space correlation characteristics of WSN data.In order to prove the improvement of this algorithm in energy consumption balance,network life-time and time complexity,this algorithm is compared with HCS,Cluster STCS,Cluster STCS?E.Cluster STRNS.The experimental results show that,in Luce data source,this algorithm improves the life-time by 155.8%,69.5%,10.0%and 66.2% respectively compared with the first four algorithms;the time complexity was reduced by 90.5%,89.2%,17.6%,95.2% respectively.