Research On Statistical Model Based Query Processing Algorithms In Wireless Sensor Networks

Large scales of deployable and inexpensive wireless sensor networks play a crucialrole in research work and applications. By querying and analyzing the data sensed bywireless sensor networks, one can effectively monitor the physical world. Sensor datasequence-based query processing has not attracted much attention in wireless sensor net-works though its counterpart has been studied extensively in time series databases. Thiskind of query can provide researchers with abundant information. For example, sensordata sequence-based query can predict the occurrences of interesting events. Wirelesssensor networks have some fundamental constraints, such as low memory storage, lowenergy and uncertainty of collected data. Thus, it is important to design effective al-gorithms to performs sequence-based queries with low energy cost, low communicationoverhead and low storage constraints.So as to overcome the inherent constraints of wireless sensor networks, this paperproposes a novel approach - HIBOR (HIstogram with Bit VectOR) to process queriesbased on data sequences collected by sensor nodes. HIBOR utilizes histograms to fit thedata distributions of sensor data sequences and bit vectors to fit the moving trends of thesequences. HIBOR partitions the whole sensor network into different clusters based onthe similarity of histograms and bit vectors. The whole sensor network logically formsa hierarchical tree after the clustering procedure. Histograms and bit vectors of sensornodes in the same cluster satisfy user specific thresholds. The hierarchical tree inherentlyforms a distributed index. Each cluster head in the hierarchical tree maintains statisti-cal information of the cluster, including average histogram, max histogram and min his-togram. HIBOR can filter query candidates at cluster heads and prune the subtrees thatcannot generate the query according to user specific thresholds. This paper process queryby example based on sequence patterns and NN-query using HIBOR approach. Exten-sive experiments on both real-world and synthetic data sets show that our approach cansignificantly reduce total communication overheads and extend the network lifetime.