基于Dirichlet分布的贝叶斯分类算法的手写数字字符识别

Handwritten digital recognition based on Dirichlet distribution and Bayesian classifier

贝叶斯分类器是一种基于概率统计的常用分类器。本文在原有的贝叶斯分类的基础上进行了改进,提出了一种基于Dirichlet分布的贝叶斯分类模型,对手写数字字符进行识别的算法。该算法用模板法进行提取特征,这种方法更易操作,提取特征的效果也好。实验表明,用贝叶斯分类方法比传统的Fisher分类算法能更好地对手写数字字符进行分类识别,且在众多领域中有较大的应用价值。

The Bayesian classifier which is based on probability statistic is a commonly used classification. This paper introduces a Bayesian classifying model based on the Dirichlet prior distribution which is improved by existed Bayesian model to recognize handwritten numbers. The arithmetic uses template method to extract feature, making operation more easily and result of extracted feature better. Experimental results show that the arithmetic is better than traditional Fisher classifying arithmetic to recognize handwritten numbers. It has been used successfully in many fields.