A knowledge acquisition technique for recognizing handprinted Chinese characters

Abstract

This paper presents a new approach to building classifiers for Chinese character recognition. A new knowledge acquisition technique is used for building incrementally a knowledge based system for the classification task. The new knowledge acquisition technique is based on ripple down rules, which is an effective method for building large knowledge bases. We extended ripple down rules to be able to acquire graphical knowledge. In this paper a Hough transform technique is used for extracting the features which are being used for a knowledge based system. A prototype has been implemented and initial experimental results are promising.