【摘要】 分析了特征维数与分类错误概率的关系 ,提出了增加新特征 ,采用混合特征提取方法进行分类器设计 .结果表明增加的新特征可以提高分类器的性能 ,并能使识别率明显提高 .在问题的概率结构完全已知的情况下 ,增添新特征不会增加Bayes风险 .更多还原

【Abstract】 This paper analyzes the relationship between dimensions of feature and the probability of classification mistakes. New features are added to increase the recognition rate. The classifier is designed by using mixed features extracting methods. The results show that the newly added features can improve the performances of the classifier obviously. And the recognition rate is higher than that it was before. If the probability structure of problem is completely known, the new features will not increase the Bayes risk.更多还原