AN ON-LINE CHINESE CHARACTER RECOGNITION SYSTEM USING COMPLEX NEURAL LOGIC NETWORKS

Abstract

This thesis is the result of the research and development of a complete on-line Chinese handwritten character recognition system. It is based on a new class of neural networks known as neural logic networks which is being researched at the Institute of Systems Science, National University of Singapore. Using the pattern recognition ability of complex neural logic networks, an implementation of a fast and accurate stroke-order independent Chinese handwritten character recognition system has been developed at the Apple-ISS Research Center suitable to be run on an Apple Macintosh personal computer occupying no more than one megabyte of RAM and one megabyte of disk storage space. To identify an unknown Chinese character written in neat, block style, the complex neural logic network has been used as a pre-classification stage to reduce the number of candidates to search. It correctly picks out two hundred candidates from a database of 3755 characters with close to 98% accuracy. From the reduced list of candidates and using a template matching approach for detailed recognition, the probability of the unknown Chinese character being identified in the top position is more than 90% while that of it appearing within a list of ten candidates is above 95%.