Text processing: robust character recognition using calibrated text and diversified feature set

Abstract

An effective algorithm for a high-performance character recognition system for printed text is presented. The system investigates characters with different aspects of the characteristics to optimize recognition performance. The research is implemented by two major phases: pattern learning and character matching. Therefore, it is not only possible to recognize characters, but also to update the database if any new pattern is detected. An initial implementation of all parts of the proposed system is reported, showing an overall recognition rate of 99.9%.<>