The Configuration of a PRD
Before the design of a PRD is discussed the structure of the finished product will be described briefly. Following the usual system description (Marill and Green 1960), a pattern recognizer consists of two parts, a receptor and a categorizer, as shown in Figure 1; preliminary processing is disregarded at this point, it will be discussed in Subsection 5. 1. The receptor may also be referred to as the feature extractor or the characteristic calculator. The categorizer is also called the decision maker or the classifier. The functions of the two parts become apparent when it is considered how a PRD (in the recognition phase) decides to which of several classes of patterns an unlabelled pattern belongs. The number of pattern classes is called Nc. In the receptor, the unlabelled pattern is exposed to a battery of different tests. The test results constitute a set of numbers γ1, γ2, ..., γq, these numbers are processed in some manner to yield the set of numbers Ξ1, Ξ2, ..., Ξp, that constitute the input to the categorizer. The p numbers may be illustrated geometrically by a point, called a pattern point, in a p-dimensional space, referred to as pattern space or decision space; the pattern point is called Ξ = (Ξ1, ..., Ξp).The p numbers are the values of the p attributes in terms of which the unlabelled pattern is now described.
KeywordsPattern Point Decision Space Pattern Class Pattern Space Pattern Recognizer
Unable to display preview. Download preview PDF.