Discrimination of sensing data in normal and abnormal situations of the monitored object or environment

The huge volume of history sensing data of a wireless sensor network need to be processed and discriminated to help the users of the data to analyze and judge the different situations of the monitored object and environment. A novel approach is proposed to first divide the history sensing data into partitions so that the data, measured when the monitored object or environment is normal, are roughly distinguishable from those measured when the object or environment is abnormal. Then the method uses a new centroid-based clustering algorithm to group the data in the partitions into different clusters. Finally the clusters of data are labeled ¿normal¿ or ¿abnormal¿ by applying the suggested heuristics.