ABSTRACT
Human beings perform the task of pattern processing in almost every instant of their working lives. The purpose of preprocessing is to suppress noise and to remove redundancies from the input pattern. The purpose of segmentation is to divide a pattern into subpatterns so that we can identify the subpatterns of interest with respect to a given application. In the decision-theoretic approach, instead of simply matching the input pattern with the templates, the classification is based on a set of feature measurements, extracted from the input pattern. "Feature-extraction" approach, where the templates are stored in terms of feature measurements and a special classification criterion is used for the classifier. Minimum-distance classifier is a linear classifier. The performance of a minimum-distance classifier is dependent upon an appropriately selected set of reference vectors. The syntactic approach to pattern recognition provides a capability for describing a large set of complex patterns using small sets of simple pattern primitives and of grammatical rules.
