基于模糊积分分类器融合的人脸识别

Face Recognition Based on Classifiers Fusion Using Fuzzy Integral

提出一种基于模糊积分分类器融合的人脸识别算法。对人脸图像进行小波变换,选取合适的小波基函数及有效的分解层数,提取低频分量系数作为分类特征设计分类器。对原图像采用2DPCA进行特征提取设计另一分类器,采用模糊积分的算法融合2个分类器并得出最终分类识别结果。实验结果表明,模糊积分能够有效融合分类器的互补信息,提高系统的分类性能,从而提高人脸识别率。

This paper proposes a face recognition algorithm based on classifiers fusion using fuzzy integral. Wavelet transform is taken on after an appropriate mother wavelet and decomposition layer are chosen by theory and simulation, and the low-frequency component coefficients of the last layer are extracted to make one classifier. 2DPCA is applied to make the other. Fuzzy integral is applied to reach the final decision. Experimental results show the algorithm can obtain complementary information and improve the classification capbility so as to promote the recognition performance.