Mobile Intelligence

27.1 INTRODUCTION

There have been several attempts to develop efficient and practical approaches to capture the different states of phenomena in the physical world using sensor networks. The most predominate model used for sensor network application involves sending sensory data to a base station for analysis [1]. Processing sensory data in the base station causes two fundamental problems. First, communicating sensory data from the physical environment to centralized servers is an expensive task in terms of energy resource consumption of sensor nodes. Second, sending streams of raw data from each sensor node to be analyzed in centralized servers may overwhelm the processing capacity of these centralized servers. In both the cases, sensory data may encounter delays that could diminish its significance.

In addition to sensing the environment, sensor nodes are capable of locally processing and wirelessly communicating sensory data. The network is, therefore, capable of in-network processing sensory data to replace the need for processing sensory data in centralized locations, that is, base stations [15]. Threshold-based techniques are one of the most commonly used techniques for event detection in sensor networks. A threshold-based technique can be as simple as a single value hardcoded into the sensor network application. An event is detected when the sensory readings exceed the threshold value. This approach, however, cannot be used to detect complex events. Generally, physical world ...