Off-Line Handwritten Chinese Character Recognition with Nonlinear Pre-classification

- In this paper, we describe a new Chinese character recognition system, in which neural networks are employed as a nonlinear pre-classifier to pre-classify similar Chinese characters, and an algorithm of clustering called Association Class Grouping algorithm (ACG) is hired to cluster similar Chinese characters. In our system, feature of contour direction is extracted to form a Bayesian classifier. Experiments have been conducted to recognize 3,755 Chinese Characters. The recognition rate is about 92%. Key words: Chinese Character Recognition, Association Class Grouping, Neural Network, Feature of Contour Direction, Bayesian Classifier.

• Type: Conference
• Year: 2000
• Where: ICMI
• Authors: Lixin Zhen, Ruiwei Dai
• Keywords: Biometrics, Chinese Character, Chinese Character Recognition, ICMI 2000, Similar Chinese Characters