Time-varying Sparse Signal Reconstruction In Wireless Sensor Networks

With the development of low-power micro sensor and wireless communication,wireless sensor networks as an intelligent system with high efficiency and fault tolerant,has been widely concerned and applied.Due to the high real-time nature of the sensor nodes in the network,the energy is limited and the communication bandwidth is small.In addition,redundant data still exist during data transmission and processing.Therefore,these factors should be taken into account when designing the state estimation algorithm for dynamic signals that applied on wireless sensor network.However,the compressive sensing theory,which is based on the Nyquist sampling theorem,solves the problem of dynamic signal estimation in wireless sensor networks.The compressive sensing theory breaks through the limitation of Nyquist sampling theorem by taking the sparsity of signal into account,and it can reconstruct the original signal from a small number of measurements with noise,which saves a lot of computational and storage resources.In this paper,we focus on the reconstruction of dynamic sparse signals in wireless sensor networks,which reads as follows:First of all,based on compressive sensing theory and information consensus filters,we propose a new time-varying sparse signal reconstruction algorithm which is also called information consensus filtering with sparse constraint.The core idea of the algorithm is the selection of the filtering: the distributed information consensus filtering algorithm,which only needs to communicate with the neighboring nodes in single hop.By combining the information filter and consensus filter,all the nodes in the network can estimate the state of the moving target of interest and reach a consensus on the state estimate.The proposed algorithm is used to reduce communication cost,enhance fault tolerance,speed up the computation and make good robustness and so on.In addition,the node in the sensor network has a limitation field on sensing the moving target state that led to the generation of the special node called ‘naive'.Nonetheless,the method can still reconstruct the signal accurately by making a good use of a small number of measurements with noise.By embedding the pseudo-measurement technique on the information consensus filtering framework,the effection of reconstructing dynamic sparse signal is more better.The simulation results show that compared with the existing algorithms,the dynamic sparse signal reconstruction method proposed in this paper is better and more effectiveness.Secondly,based on the existing dynamic sparse signal reconstruction algorithm,a centralized Kalman filter time-sparse signal reconstruction algorithm based on variational Bayesian is proposed for the case where the measurement noise is unknown and slowly changing.The system model of the algorithm is susceptible to external disturbance,which causes the observed noise parameter to change,and the performance of reconstruction is reduced.Therefore,the solution is to use the variational Bayesian method to estimate the random variables which are composed of the signal state and the time-variant measurement noise.And then,measurements noise variances were approximated by variational Bayes,thereafter,system states were updated by the centralized Kalman filter with the state space sparsely constraints.All above steps can achieve the dynamic sparse signal precision reconstruction.The simulation results show that the centralized Kalman sparse signal reconstruction algorithm based on variational Bayesian can achieve better tracking of the measurement noise variance.Compared with the standard centralized Kalman filter method without the noise adaptivity,the state estimation of the proposed method is more accurate.