基于机器学习的手写汉字识别的研究

MACHINE LEARNING BASED HAND-WRITTEN CHINESE CHARACTER RECOGNITION AND ITS IMPLEMENTATION

本文实现了一种基于机器学习的手写汉字识别方法.针对手写汉字的特点,选择并提取了横竖笔划特征、周边特征、结构划分特征、分区特征点、黑点重量等作为分类特征.在分类策略中采取了先粗分类后细分类的多级分类方法.并将决策树算法ID3成功地应用到分类策略中,在识别时中利用决策树引导特征提取,减少了特征提取的数量,从而大大提高了识别速度.

A method of handwritten Chinese character recognition based on machine learning is presented and implemented. According to the characteristics of chinese character. Some features such as horizontal-vertical stroke, contour, area pints, structure partition, black point weight are selected and extracted as the classification features. Then in classification, strategy, a multiple layer classification method in the order of pro-classification and fine classification is applied. And the induction of decision trees algorithm ID3 is successfully applied into classification. It guides feature extraction in the procedure of recognition and subtracts the number of extracted features, hence it accelerates recognition speed greatly.