Syntactic Pattern Recognition *
K. S. Fu · Pattern Recognition · 2019
The many different mathematical techniques used to solve pattern recognition problems may be grouped into two general approaches. They are the decision-theoretic (or discriminant) approach and the syntactic (or structural) approach. This chapter briefly reviews the recent progress in syntactic pattern recognition and some of its applications. In syntactic methods, a pattern is represented by a sentence in a language which is specified by a grammar. The language which provides the structural description of patterns, in terms of a set of pattern primitives and their composition relations, is sometimes called the “pattern description language”. Since pattern primitives are the basic components of a pattern, presumably they are easy to recognize. The chapter presents a brief introduction to tree grammars and their application to syntactic pattern recognition. In order to describe noisy and distorted patterns under ambiguous situations, the use of stochastic languages has been suggested. With probabilities associated with grammar rules, a stochastic grammar generates sentences with a probability distribution.