Research On The Applycation Of Compresive Sensing To Wireless Sensor Network

Wireless sensor network (WSN) is the extension of the existing network to the physical world, and it is also one of the most important part of the Internet of things (IOT). WSN has been widely used in environment,military, medical and other fields. WSN, network technology and computer technology constitute the three pillars of information technology. But limited resources of nodes lead to a lot of difficult problems in some aspects such as the scale of deployment and life time, which greatly influences its practical application. Compressed sensing (CS) is the emerging theory of signal processing, and overturns the traditional Nyquist criterion. And CS method can transfer the complex operation from coding side to the decoder side, conforming to the resource constraints characteristic of WSN. At present, the CS technology is widely applied in many fields, mainly used in data collection, positioning and routing in WSN.Based on the CS theory, this thesis will discuss its application in the WSN, and combine the characters of CS with the structure of WSN and the correlation between sensory data to solve the problem that nodes consume energy too quickly because of the unequal node load or the great data redundancy in the process of data collection and distributed data storage.In this way, energy efficiency will be improved in the entire network.Firstly, it's known that the combination of the existent hybrid CS method and traditional clustering methods fails to realize the efficient utilization of energy, because the data redundancy is still large. For this problem, an analytical model of cellular clustering is put forward to study how the special hexagon structure can be combined with CS for a better performance in energy consumption. Then, on the basis of hexagon clustering model, a new method of hybrid CS is presented, which performs better on power consumption than other hybrid CS. Extensive simulations confirm that our method can reduce energy consumption significantly.Distributed data storage (DDS) provides a reliable method for the whole network data recovery for WSN, and can be realized just by visiting only a small part of the sensor nodes. Traditional DDS schemes only consider the geographical spatial correlation of the perception data between adjacent nodes, while ignoring the time correlation between multiple consecutive time slots. This makes the data redundancy has not been fully eliminated. This thesis proposes a new scheme of DDS called CSSTDS,which takes both the space correlation and time correlation into consideration at the same time, and compresses data from two dimensions.This scheme realizes the improvement of energy efficiency by decreasing the number of transmitting and receiving packets. Finally kronecker product is used to process two-dimensional CS problems, and reconstruct the original signal.