Research Of Data Compression Method Based On Compressive Sensing In Wireless Sensor Networks

Analog sources of information and digital information processing tools are the two main features of the digital information age. Signal sampling is the necessary bridge of these two features. However, with growing demand for information and the bandwidth to carry information continuously widening, making the sampling theorem of existing is more and more difficult to broadband signal processing. And brought great pressure to signal acquisition, storage, transmission, processing. For example, in sensor networks, with the increasing number of nodes, network functions continue to strengthen, the transfer and storage of huge amount of data is a difficult task.This thesis describes the theory of compressed sensing and a wireless sensor network data processing theory. On this basis, the improved method of compression Sensing is introduced into the data acquisition process of wireless sensor network, that is sparse collect data and observe at the sensor nodes, making the network only need transfer and storage a small amount of data, at the same time use the designed reconstruction matrix to reconstruct the original data at the end of receiving.Using a sensor array which integrate temperature, humidity and gas sensors as a network node, its measured data as the data to be processed, to sparse and observe, the observed data less than 20% of the original, the error of reconstructed data is 1.5011. And reconstruct and sparse data only for a node formed by sensor array consisting of gas sensors, and then use the BP neural network BP neural network to the quantitative and qualitative identification of gas of the reconstructed data, to find smallest sparse coefficient m which make CO, H2, CH4 to qualitatively identify, and then look for the minimum sparse m which make the accuracy of quantitative detection of CH4 can reach 80%. Spatial correlation between the sensor nodes can be exploited in order to compress and reconstruct sensor observations in an energy efficient manner based on coding/decoding algorithm of compressed sensing. Finally, the analysis of relationship energy efficient performance of compressed sensing are carried out in simulation. Simulation results show that compressed sensing can achieve acceptable estimation accuracy in an energy efficient way.