ABSTRACT
The intensive development of wireless technologies and the increasing miniaturization of RF devices and micro electro-mechanical systems (MEMS) have been driving forces in the advancement of small and tiny computing devices, such as WSN technology. These devices are inter-connected and form a computational network that is capable of providing a frontline processing scheme for applications such as event detection and remote monitoring. Because this type of network has a large number of computing nodes that have limited power, storage, and processing capabilities, it is referred to as a fine-grained network. The ability to acquire resource-awareness characteristics was discussed in
the previous chapter and is essential for the design of distributed applications, including pattern recognition. The DHGN as a distributed pattern recognition scheme was developed with adaptive granularity characteristics built into its design. A distributed and parallel pattern recognition scheme for applications in fine-grained systems was introduced by Khan and Mihailescu [2]. In a fully distributed configuration, each GN is assigned to a single compute
node, and the collaborations of inter-connected compute nodes form a DHGN subnet. The simple bias array search computations involved for each node make this configuration well-suited for fine-grained networks that have limited processing and storage capabilities, such as a WSN. We will demonstrate the robustness and scalability of the DHGN for a distributed recognition process over a fine-grained network using a number of DPR applications.
