Databases Reference
In-Depth Information
FIGURE 7.2: DHGN multi-feature recognition scheme, a collection of DHGN
networks that analyze patterns using multiple sets of features.
recognition can be extended provided su
cient computational resources are
available. The number of features, f, is directly proportional to the computa-
tional resources available for the recognition scheme, c; f ∝ c. These resources
are in the form of distributed computational networks, which provide greater
scalability for recognition purposes.
7.2.1 Conceptual Design and Implementation
The design for multi-feature recognition comprises a collection of DHGN
networks. Each network performs a distributed recognition scheme for a single
feature. Figure 7.2 presents the DHGN multi-feature recognition approach.
In this configuration, a coordinator node is used for data acquisition and
networks coordination. This node communicates the patterns received to the
SI module node on each DHGN network. Each SI module has a copy of the
pattern set for the recognition process. The SI module starts the recognition
process by generating a single feature obtained from the input patterns. The
feature data are used as a pattern for recognition purposes. The rest of the
recognition procedures in each network are similar to the original DHGN
scheme. The results for each recognition process conducted by each DHGN
network are sent to its respective SI module. Each SI module produces a result
for the recognition/classification of each pattern in context with the operator-
specific accuracy parameter(s). These parameters can include recall, precision,
and error values. The results are passed to coordinator node, and the error is
