Mobile Intelligence

27.3 VOTING GRAPH NEURON APPROACH

In this section, the VGN approach is proposed. The VGN approach recognizes events by utilizing a novel energy-efficient template-matching approach. In essence, the VGN approach uses graph neuron (GN) concepts [11–13]. The VGN represents sensory data in patterns that are used for detecting events of interest. Sensor nodes are depicted as memory cells that can store subpatterns. The VGN approach is concerned with how to infer global information from the local information of sensor nodes. Patterns are matched by means of votes. Contrary to the GN algorithm, which uses a simple binary decision, the votes of sensor nodes are the units of collaborations, list all possible reference patterns matching the unknown patterns, and are designed to allow sensor nodes to exchange decisions to reach a network decision. Additionally, node collaboration, distributed in-network processing, distributed storage, and life-cycle management are employed to enable practical and energy-efficient pattern-recognition using resource-constrained sensor nodes.

27.3.1 VGN Overview

Performing event recognition in sensor networks is a challenging problem. Events need to be recognized using geographically distributed and ad hoc infrastructures. The processing of information needs to be conducted within the network using very small, resource-constrained sensor nodes. The sensor nodes wirelessly communicate using a shared wireless channel. The channel's bandwidth and its varying quality ...