Mobile Intelligence

27.5 COMPARISON

The VGN approach can be compared to the GN algorithm. Nonetheless, the GN algorithm is a neural network-based approach in which sensor nodes are mapped into a neural network to store and recognize patterns in a distributed manner. The GN algorithm recognizes events by local node processing and communication. The GN algorithm does not incorporate any sleep mode-based strategy, and thus all nodes remain active all the time. Each GN node processes subpatterns (pairs) to produce a boolean result indicating a match or no match to the unknown pattern. Although to recognize any pattern, positive recognition results need to be collected from all the nodes, the GN algorithm's recognition accuracy is very low due to pattern interference. In the existence of pattern interference, similar to shared areas discussed in Figure 27.2, boolean local results do not provide enough information to recognize patterns. To improve the GN algorithm's accuracy, the set of reference patterns are required to be selected carefully to ensure that the stored pattern set is an interference-free set—which may not be always feasible. On the contrary, the VGN approach is not affected by pattern interference, since the node collaboration and vote set exchange provide sufficient information to resolve interferences and match patterns correctly.

In regard to the communication requirement, the GN algorithm depends solely on local collaboration to recognize patterns. Each GN node exchanges its local results ...