A simple learning decision algorithm for character recognition and pattern classification

A learning decision algorithm, using a set of distinctive features, is described and applied to character recognition. It is based on assumptions which have a wide application area. Emphasis is given on the help it can provide to an industrial user who has to design a character recognizer. In relation to a statistical model, convergence properties are proved at a theoretical level. At a practical level, satisfactory results are shown for three different applications. Tools for performance analysis are sketched.