目标回声分类特征的冗余性评价

Redundancy evaluation of the target echo classification feature

对目标回声分类特征的冗余性进行评价与分析。对于2个特征维之间的冗余性,采用基于特征间线性相关系数作为冗余性度量,通过特征的相关系数矩阵分析了特征维之间的冗余度。此外,本文提出了基于特征协方差矩阵近零特征值的冗余性度量,据此可以进一步分析特征组内部的多维组合冗余。最后,利用以上2种冗余性度量,对5种目标回声分类特征各自的冗余性以及它们合并后总特征的冗余性进行了评价与分析,明确了其中的冗余关系,为目标回声分类特征的优选和组合应用提供了重要依据。

The redundancy evaluation and analysis of target echo classification feature set is studied. For the redundancy between two features, the linear correlation coefficient is used as the redundancy measure. And the feature redundancy can be analyzed via its correlation matrix. Moreover, the paper puts forward another redundancy measure which is based on the near-zero eigenvalue of the features covariance matrix. By this, we can further analyze the multi-feature redundancy within the feature subset. Then, the two before-mentioned redundancy measure are used to evaluate and analyze the redundancy of each five target echo classification feature set and the total features. And the redundancy situation between them is made clear,which provides important information for the selection and combination of the target echo classification feature set.