摘要
特征选择是水声目标识别领域的重要环节之一。提出基于互信息的顺序向前特征选择算法,通过计算特征之间的互信息和特征与类别间的互信息对所有特征的分类能力进行排序。提取了实测4类水声目标进行特征选择和分类实验,结果表明:该算法能够选择有效特征子集,得到较高的正确识别率,并且运行速度快,稳定性强。
Feature selection is important in classifying underwater acoustic target. In this paper, an algorithm of forward order feature selection based on mutual information（SFFSMI algorithm） is proposed. This algorithm sorts the classifying abilities of all features by calculating mutual information between different features and the mutual information between features and classes. The features of 4 classes of underwater targets are extracted and used for feature selection and classification experiment. Experimental results show that, this algorithm can choose effective feature subsets with high correct identification rate and it runs fast with high stability.
出处
《声学技术》
CSCD
2014年第4期359-362,共4页
Technical Acoustics
基金
水下测控重点实验室基金(9140C2)
关键词
水声目标识别
特征选择
互信息
排序
underwater acoustic target classification
feature selection
mutual information
sorting
