Sparse Representation for Feature Selection in face demographic classification

作者：Yu Chanjuan; Fang Yuchun^*

来源：Journal of Computational Information Systems, 2014, 10(24): 10635-10642.

DOI：10.12733/jcis12652

• Demographic information
• Face Demographic Recognition
• Feature combination
• Feature selection methods
• Human face image
• Rich features
• Sparse representation
• Training strategy

Since the valuable demographic information like gender and race involved in the human face images,and the importance of feature selection, in this paper, we explore the feature characteristics and training strategies of Sparse Representation for Feature Selection (SRFS) and apply it to face demographic classification. The proposed SRFS mainly use sparse values to select the most effective and semantically rich features at feature selection stage. And we try to use the largest part and the smallest part of sparse values to perform the feature selection. We also propose two strategies SRFS1 and SRFS2 to obtain the sparse values. In experiments, we find a more effective feature combination obtained by ascending sparse value and compare the training strategies. On very large face database MORPH-II, the SRFS proves to potentially be an excellent method applied in feature selection for face demographic classification by comparing with some other typical feature selection methods.

• 出版日期2014
• 单位上海大学