The Use of Negative Printing and Attributed-Token State Modelin Character Recognition

This article proposed a new scheme for presenting and recognizing of printed character. The method employs the concept of negative painting i.e. uses surrounding pieces of information as feature elements instead of analyzing the target character directly. Curvature continuity is first parameterize as one of six symbolic consistencies (feature primitives). Then, surrounding pieces are represented as attributed-token state sequences (ATS-sequences) of symbolic consistencies. The advantage of the proposed ATS-sequences is that it converts characters into token-sequences without loosing their structural uniqueness wherein the classification can be done by a computational hierarchical classifier.