查看文章

This paper presents an off-line handwritten, isolated character recognizer based on the artificial immune system (AIS) and qualitative rules-based system (QRBS). AIS acts as an optimizer. It selects the best candidates for training. Each candidate is used from several character features previously selected based on some structural and statistical techniques. QRBS works as a qualitative recognizer system. It utilizes qualitative rules to recognize characters. It handles both imprecision and uncertainty in handwriting during the training and classification phases that make some characters unreadable and may decrease the accuracy of the overall process. Experiments are conducted on the MNIST and English letter databases. Comparisons with other recent approaches using the same database indicate that this approach is effective.