A Method of Feature Selection Based on Modified FDR

摘要：

Feature selection plays an important role in the research of Pattern Recognition. Because of the Fisher Discriminant Ratio cannot denote the separability of the feature component comprehensively, this paper proposed a modified Fisher Discriminant Ratio based on the entropy theory. It denotes the distributing information among the different sorts of targets. And the feature selection method uses the modified FDA. Simulation result of radar HRRP target recognition shows that the new method has a good performance.

关键词： features selection FDR entropy

作者: Ben Cao D.-B. Ma K.-F. Zhang C.-Y. Niu

作者单位: Department of Information Science, Zhengzhou Information Science and Technology Institute, Zhengzhou, Henan, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 611-613

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）