Density Based Multisensor Data Fusion for Multiapplication Wireless Sensor Networks

The trend in the design of wireless sensor networks (WSNs) has recently shifted from an application-specific approach to a virtualization and resource-sharing view. Current WSNs allow the sensing and communication infrastructure to be shared among multiple applications thus optimizing the use of resources. However, as the number of applications sharing the WSN infrastructure grows, storing and analyzing requirements of all applications at the network level becomes unfeasible. To tackle this challenge, in this paper we proposed a multisensor data fusion method that requires no knowledge of the application specific requirements. For performing the analysis of the data, in our proposal we leverage concepts of the area of pattern recognition to create abstract sensors based on intervals in the monitored data. Considering that sensed data have a measuring range, given by their maximum and minimum values, our proposed methods group the data samples in abstract sensors based on density and peaks on the dataset. Both real and simulated experiments have shown that our proposal is accurate and reduce energy consumption.