Research On Compressed Sensing Based On Ratio Difference Sparseness In Wireless Sensor Network

Abstract:Due to the limitation of the energy supply and communication bandwidth, efficient data compression is needed for wireless sensor networks. Compared with traditional data compression method, compressed sensing is considered to be an efficient data compression method because its sampling rate is much lower than the Shannon-Nyquist sampling theorem. Compressed sensing requires the data to be sparse, however, the data obtained by the actual sensor network is not sparse, but it has a certain spatio-temporal correlation. Therefore, research on compressed sensing through sparse processing based on the correlation of sensor data is very important in theory and practice.Based on the spatial and temporal correlation of the signal collected by sensor nodes, four different kinds of sparse models and corresponding compressed sensing algorithms are presented.Firstly, two kinds of distributed compression sensing algorithms towards non-sparse signal for clustering network are proposed. These two models are based on the stable ratio and stable incremental of nodes output respectively. Experimental original test signal composes of a sinusoidal signal, a linear signal and a constant signal. And orthogonal matching pursuit algorithm is used to reconstruct the signal affected by noise. The experimental results show that these algorithms can restore the signal accurately in the case of reducing the number of the measured value and suit for energy-efficient distributed compression.Secondly, considering the spatial and temporal correlation of images collected in continuous time for the video sensor node, we propose an image code algorithm based on the stable ratio of pixels of two adjacent frames. At first calculate the gray ratio between each column element and the first column separately of adjacent two frames, and then calculate the difference between the two ratio matrixes and then set a predetermined threshold value to make it a sparse matrix. In the end, divide the matrix into blocks and use compressed sensing to encode them and use the same decoding method to reconstruct the image at the receiving end. The experimental results show that the model can obtain a more remarkable reconstruction under the condition that the brightness of the scene is changed. As the lower sampling rate of compressed sensing, this algorithm can greatly reduce the demand for node storage and the power to compute, and extend the life cycle of the network.Finally, under the same construct conditions with the image code algorithm based on stable ratio between pixels of adjacent frame, we propose another block compressed sensing model based on the stable incremental of pixels. At first calculate the difference of each column element with the preceding column elements, and then calculate the ratio of the difference and the first column, and next seek the different ratio between the adjacent two frames. Set a threshold to make the matrix become sparse. Use block compressed sensing to encode the image and at the receiving end, use the same reconstruction algorithm to rebuild the image at sink node. The experimental results show that no matter how the brightness of the scene changes. This method can still restore the image with high possibility and its reconstruction quality is much higher than the method that just does a differential processing and blocking of the image.The result of this paper shows that it can promote the application of compressed sensing and reduce energy consumption of wireless sensor networks through data compression.19figures,6tables,56references.