The Research On Wavelet Based Data Processing & Clustering Algorithms In Wireless Sensor Networks

With the rapid developments of MEMS (micro-electro-mechanism system), integrated circuits, wireless communication and information networks, wireless sensor networks (WSN) are becoming available. In addition to data collecting, processing and communicating, sensor networks are also inexpensive and easy to be placed, so they are highly researchful and useful. Recently, WSNs have attracted more and more concentration from both academia and industrial community, and have been applied in many fields. Energy is a key resource for sensor networks, which is dominated by the energy consumption of data transmission. So, how the data is routed and processed in network are key problems to be solved. The research on clustering and wavelet-based data processing is the topic of this thesis.(1) The concept, architecture, character and current popular application of wireless sensor networks are introduced, and the background and significance of the topic are explained also. Then, some typical clustering routing protocols are analyzed, such as LEACH, ACE, and the theory of wavelet is introduced to provide theoretical and experimental basis for the research in the following chapters.(2) Aiming at wavelet data compression, a feed-back clustering algorithm based on coarse data correlation in sensor network is proposed. The clusters whose data have better correlations are formed by employing the coarse correlation among sensing data, then some reconstructed data fed back from Sink are compared with the real data in the corresponding sensors and the optimized clusters are obtained. Theoretically and experimentally, we show the proposed algorithm can make the wavelet data compression in sensor networks more efficient, e.g. less error, higher compression ratio. Moreover, it can prevent the data submergence in wavelet data compression.(3) On the basis of feed-back clustering algorithm by using coarse data correlation, a new cluster head rotation strategy based on Master-Vice cluster heads is proposed. For sensor networks, the life is very important for the performance. Aiming at how to prolong the life of sensor networks, the idea of Master and Vice heads is proposed. An operating mechanism of Master-vice cluster heads is designed to avoid"fake death"caused by the failure of heads and prolong the life of networks. Experimental results prove the new strategy is effective. (4) Aiming at the irregurlar wavelet data processing in sensor networks, a new irregular data processing algorithm based on variable Voronoi polygon is proposed to gain accurate data and decrease the energy consumption for data transmission further. Corresponding to the changes'levels of sampling data, Voronoi regions can be adjusted according to the importance of data. Meanwhile, based on the scale coefficients and the precisions, Sink makes a decision whether part of wavelet coefficients are needed for reconstruction. The experiments show that the new algorithm can reconstruct data precisely, and avoid data submergences in irregular wavelet transformation efficiently, and prolong the life of network effectively.