Research And Application Of Median Filtering Algorithm For Wireless Sensore Network

Information processing is the central task of WSN applications. As a very important technique, digital filters are widely used in many fields of information processing. The digital filters for WSN applications must be simple as well as high efficient to meet the performance demands of WSN because the sensor nodes are often highly constrained by the limited power supply, computing capability and memory size. In principle, median filters are such a kind of filters which are suitable for WSN applications. But, median filters are not perfect and most of them are used for image processing. So it is important to improve the median filters for non-image information processing of WSN in theoretical study and practical applicationAccording to the characteristics of limited resource in WSN, this thesis first compares and analyzes the existing median filters used for processing non-image signal. Then, three problems to be solved are summarized:noise suppression capability, calculating efficiency and frequency selection capability. Hence, an improved pseudo-median filter (IPMF) is suggested to enhance the impulse suppressing capability of original pseudo-median filter by adopting an impulse detection procedure. Also, the IPMF maintains a relatively high efficiency in calculation. Further more a combined median filter (CMF) which combines the advantages of pseudo-median filter and standard median filter is proposed to smooth impulses much better than IPMF. For frequency filtering applications, the thesis designs a kind of negative weighted repeated pseudo-median filter (NWRPMF) and its simplification. The simulations in Matlab show that the frequency selection characteristics of NWRPMF are much better than the existing negative weighted pseudo-median filter (NWPMF).After comparing and analyzing of the typical design of WSN node, a test platform is build up based on MSP430F1611 MCU and IAR instrument. Then, the three proposed improved algorithm, IPMF, CMF and NWRPMF, are verified by pulse signal flitting experiments on the platform. Both the simulations and experiments reveal that the three proposed algorithms which satisfy the requirements of signal processing in WSN node can be widely applied to process vibration, pulse, temperature and many other non-image signals. So the proposed algorithms will have wide applications.