Lightweight Event Detection Scheme using Distributed Hierarchical Graph Neuron in Wireless Sensor Networks 1 x Lightweight Event Detection Scheme using Distributed Hierarchical Graph Neuron in Wireless Sensor Networks Asad I. Khan, Anang Hudaya Muhamad Amin and Raja Azlina Raja Mahmood Monash University Australia 1. Introduction Existing breakthrough in communication technologies have lead to the rapid growth of emerging networks in particular the wireless sensor networks (WSNs). These networks emerged from the confluence of wireless communication technology, extensive computational schemes, and advanced sensor technology. WSNs are created from a collection of self-organised wireless and battery-powered devices with sensing capabilities. The future of this kind of networks is promising, as been mentioned by (Stankovic, 2008), “The potential of these systems is nothing short of revolutionary. This technology will affect all aspects of our lives, bringing about substantial improvements in a broad spectrum of modern technologies ranging from healthcare to military surveillance”. The current scenario of WSN deployment is however is still far away from its fullest potential. To date, WSN has only been demonstrated for humble applications such as meter reading in buildings and basic form of ecological monitoring. In order to achieve its fullest potential, WSN requires an intelligent computational scheme which at present is still lacking. Common approach implemented within existing WSN applications usually involve a number of processing steps including sensory data capture and conveyance of these data to a central entity known as the base station for further refinement and analysis. Consequently, this approach would lead to a system bottleneck, if it is scaled up for widespread use. Furthermore, processing delays would intermittently occur due to the high latency between data capture/aggregation and processing time. These limitations make WSN less suitable for real-time monitoring applications. We require a new approach for an improved data processing within WSN that has the abilities to process sensory data in situ within decentralised manner and to generate highly condensed and sophisticated outputs internally. These abilities will alleviate the bottleneck problem within WSN through on-site computations, and improve the detection performance by reducing the processing delays. In this chapter, we will describe a lightweight and distributed event detection scheme within WSN infrastructure with one-shot learning pattern recognition capability. This 1