融合过滤和封装方式的特征选择算法

摘要

已有特征选择算法不能有效降低特征维数,且稳定性较低.为此,提出一种融合过滤和封装方式的特征选择算法.在封装式算法中,设计能保持图像之间拓扑结构的特征选择判据,在过滤式算法中,以Fisher Score为判据,采用单独最优的特征搜索策略.实验结果表明,将算法应用于人脸识别中,能提高识别率,降低特征维数,且具有较好的稳定性.%The feature selection algorithm can not effectively reduce the feature dimension, and the stability is lower. In order to solve this problem, this paper proposes a feature selection algorithm of fusing filtering and packaging mode. In packaging type algorithm, it designs the feature selection criterion which can maintain topological structure between image, uses Fisher Score as criterion in the filtering type algorithm, and the individual optimum search strategy is used in this paper. Experimental results show that this algorithm can improve recognition rate, reduces the feature dimension, and has good stability in face recognition application.