A neural network expert system for Chinese handwriting-based writer identification

Some techniques have been reported for Chinese off-line handwriting-based writer identification and signature verification. Some of them assume that the written text is fixed, and others presented don´t need the presupposition since they identify the writer of a piece of handwriting text from the global style. But most of them are not used in practical application and have some disadvantages, which cannot be solved well now. We present a system integrating an expert system and a neural network to identify the writer of Chinese handwritten text. Designed for a practical application of writer identification, this system can perform as well as a human expert and exhibit characteristics of a traditional symbolic expert system with high accuracy.