Research On Data Compression Algorithm For Wireless Sensor Networks

As a new network technology, Wireless Sensor Networks (WSNs) combine thelogical information world with the objective physical world, and change theinteractive way between human beings and nature, thus it is widely used in industrialcontrol, medical care, military monitoring and intelligent traffic, etc. However, thedevelopment of WSNs faces a lot of problems that need to be sloved urgently due toits specificities. In WSNs, there is a large amount of redundancy in the collctedone-dimension data, including the temporal redundancy in the data acquired by thesame sensor node and the spatial redundancy in the data acquired by the adjacentsensor nodes, and the redundancy among the data of different measurements. Inaddition, there is also a large amount of redundancy in the collected two-dimensionimage duo to the strong correlation between adjacent pixels. The computing ability,storage ability and battery power of sensor nodes are restricted severely, therefore, ifthe original data with redundancy is transmitted directly, it would lead to a waste ofcommunication bandwidth and energy consumption, and ultimately affect the lifetimeof sensor network.Considering the redundancy in the collected on-dimension data andtwo-dimension image of WSNs, three novel data compression algorithm for WSNsare proposed in this paper.(1) Considering the temporal and spatial correlation among the collectedone-dimension data, a data compression method for wireless sensor networks basedon adaptive optimal zero suppression (AOZS) is proposed. The AOZS algirhtm sortsthe collected data in ascending order in sensor nodes and cluster-head nodesrespectively, and the correlative bit and digital factor is analysed to find an optimaldigital factor, finally the AOZS algorithm performs zero suppression and encodeoperation base on the optimal digital factor, thus removing the temporal and spatialredundancy, decreasing the data transmission in the network.(2) Considering the correlations both within the values of a single measurementas well as among the values of different measurements, and the mutability that mayexist in collected data, a wireless sensor network data compression algorithm basedon outlier screening and piecewise polynomial regression (OPR) is proposed. The OPR algorithm first chooses a base signal according to the distribution characteristicsof the collected data, and then implements polynomial regression to approximate thepiecewise data sequence to the base signal, meanwhile, the possible outliers aredetected and screened. Thus ensuring the data accuracy can satisfy the system’srequirements, and decreasing the data transmission and energy consumptioneffectively.(3) Considering the strong correlation among the two-dimension image, animage compression algorithm based on in-cluster distributed processing for wirelessmultimedia sensor networks (ICDP) is proposed. The ICDP algorithm selects aplurality of auxiliary sensor nodes in the clusters by using the energy priorityselection principle, and then the computing process of multi-level wavelet transformand encoding operation are distributed to the auxiliary nodes, so as to balance theenergy consumption of each sensor nodes and prolong the lifetime of sensor network.The above three algorithms are geared to the needs of one-dimension data withsingle attribute, one-dimension data with different attributes and two-dimensionimage respectively. These works aim to decrease the energy consumption, reduce thenetwork delay and prolong the network lifetime. In this paper, MATLAB simulationexperiments are done to analyse the performance of the above three algorithms, andour simulation results demonstrate that all three algorithms can ensure the dataaccuracy meets the system’s requirements, compressing thier respective researchobject effectively, decreasing the energy consumption and proloning the networklifetime.