Sensor Network-aware Data Compression Technology

Recent advances in wireless and embedded technique have enabled the development of low energy consumed sensor nodes. In sensor network, the disseminating of data will cost much energy in wireless sensor network. Under the circumstance of guaranteeing high data precision, how to reduce the data exchange efficiently is an important research problem.In this paper, we propose four compressing methods according to the different data characteristic of sensor network.The method based on time serials compresses the whole time serials into many segments, which are expressed by a constant value and the end-piont of a segment, which can obtain well effect on the data changes slowly. The method based on the Gaussians distribution constructs compressed snapshot for data and compresses data from continues periods into a compressed structure, which is especially efficient to sensor data obeyes Gaussians distribution. The method based on wavelets specifies wavelet coefficients can be compressed using a tree structure and minimizes coefficients greatly guaranteeing the data precision,which is not only efficient for data that fluctuates lessly, but also for data that exists abnormity. The method based on data correlation extracts the correlation informations between data of different attributes or data from different intervals which summarizes the data characteristic and forms a base signal, then encodes data with regression parameters and base signal, the stronger of the data correlation, the better of its performance.The theory analysis and the result of experiments demonstrate that our methods are efficient in compressing the sensor data in sensor network. We can choose different compressing method according to the characteristic of sensor data.