A Programmed Algorithm for Designing Multifont Character Recognition Logics

Abstract

This paper contains a discussion of a programmed algorithm for designing multifont recognition logics. This design algorithm automatically derives effective recognition logics from representative samples of characters. It has been applied to design. translational invariant logics for upper and lower case typewritten alphabets. Recognition performances in the order of a few errors per thousand characters have been achieved for mixed-font characters from six to 11 different fonts over a wide range of printing quality.