A Trainable Syntactic Model for Syntax Specification and Recognition of Hand-Drawn Two-Dimensional Patterns

摘要

In order to make direct graphic input possible, which provides a most natural interface between a user and a computer, suitable pattern recognition schemes become essential. Schemes that employ feature extraction for recognition are almost invariably problem-oriented and thus lack generality. This dissertation presents the details of a syntactic model for on-line, real-time recognition of handdrawn two-dimensional patterns. The proposed model can handle different classes of patterns in a uniform manner. Theory and algorithms of a grammar-building system are presented. The details of a table-driven analyzer are described. (Author)