关键词: 代价敏感, 主成分分析, 人脸识别, k最近邻

Abstract: Existing face recognition algorithms aim to achieve high recognition accuracy, implicitly assuming that all misclassifications lead to the same losses. This assumption, however, may not hold in the practical face recognition systems. Motivated by this concern, a new face recognition approach based on Cost-Sensitive Principal Component Analysis （Cost-Sensitive PCA） is proposed in this paper. It incorporates a cost sensitive function into Principal Component Analysis theory and determines the different loss cost by differentiating losses caused by different error recognition, which achieves more accurate face recognition. The experimental results on AR, FERET and UMIST face databases show that the proposed method achieves higher k nearest neighbor classification recognition accuracy with the least cost compared with the classical subspace-based face recognition methods.

Key words: Cost-Sensitive, Principal Component Analysis（PCA）, face recognition, k Nearest Neighbor（kNN）