Databases Reference
In-Depth Information
FIGURE 9.14: Damage detection using the DHGN distributed pattern recog-
nition scheme. (This figure is a copyright of and reproduced with permission
of Civil-Comp Ltd. Previously published in [98].)
for event occurrences in a resource-constrained network, such as a WSN. There
are several benefits to the DPR implementation. The distributed approach
uses a simple bias array representation that offers low memory consumption
for event data storage. Furthermore, the recognition scheme only stores sub-
patterns/patterns that relate to normal events rather than keeping the records
of all occurring events. This work also demonstrated that this new approach is
most effective for small subpattern sizes because it uses only a small portion of
the memory space of a typical physical sensor node in a WSN. In addition to
this e
cient memory usage, DPR schemes, such as the DHGN, eliminate the
need for complex computations for event classification techniques. With the
adoption of single-cycle learning and adjacency comparison approaches, the
DHGN implements non-iterative and lightweight computational mechanisms
for event recognition and classification. The distributed characteristic of the
DHGN implies that it is readily deployable over a distributed network. With
such features, the DHGN can perform as a front-end detection scheme for
event detection in a WSN. Through a divide-and-distribute approach, com-
plex events are perceived as a composition of events occurring at a specific
time and location. By incorporating a spatio-temporal evaluation of events,
this new approach would be able to be used in event tracking in the future.
