Information Fusion for Face Recognition Using Fuzzy Integral
AI Xue-yi · Journal of Chinese Computer Systems · 2008
This paper proposes a face recognition approach to combine the global feature and local feature of human face and aggregate the resulting individual classifiers using fuzzy integral. The present face recognition system is not accurate enough to acclimatize itself to the uncontrolled environments. First, the paper produces essential facial features such as eyes, nose, and mouth, so the cropped image and local facial features are obtained. In the sequel, fisherface method is used to extract the features and reduce the dimensionalities of those features. Then three classifiers based upon local facial images and a single classifier based upon global face images are constructed. Finally, the information fusion of the individual classifier is realized through the Choquet fuzzy integral.Experimental results show the proposed system obtains complementary results and improves recognition performance substantially.