Stroke-based intelligent character recognition using a deterministic finite automaton
D. Alvarez, Ramón Ángel Fernández Díaz, Lídia Sánchez-González · Logic Journal of IGPL · 2015
This article introduces a methodology for intelligent character recognition based on a new feature extraction technique that relies on the number of vertical strokes and the relative positions of the horizontal strokes connected to them. We also propose a new grammar in order to feed these features into a Deterministic Finite Automaton (DFA) that will generate string representations of every character. For a single writer, a set of training characters let us generate a XML-formatted Knowledge Base that, in combination with an Inference Engine, will provide a system to identify unknown characters' string representations. Our approach allows the construction of Knowledge Bases and their use with other writers without any previous training. The experiments with several writers and, hence, several Knowledge Bases, show promising performance rates, especially for the recognition of characters written by the owner of the training set.