Recognition of handwritten Chinese characters by multi-stage neural network classifiers
Hsin-Chia Fu, Kuo-Ping Chiang · 2002
This paper presents a multi-stage neural network classifiers for handwritten Chinese character recognition. In the proposed system, the authors have developed: (1) a two stage recognition structure: (a) an overlapped c-means clustering algorithm to implement a coarse classifier, (b) a Bayesian decision based neural network as a fine classifier, (2) feature selection and reduction methods, (3) a recognition system on a personal computer, which requires only 3.98 MB RAM for feature vectors storage. By using a large database (5401 characters/spl times/100 samples), the training and testing results show the efficiency (recognition time: 0.885 second per character on a Pentium based PC) and robustness (recognition rate: 86.68% and 93.60% of top one and top three respectively) of the proposed system.