Face Recognition based on Independent Component Analysis and Core Vector Machines

In the process of face recognition, ICA (independent component analysis) is used to extract face feature, which is the coefficients of face images projecting to the independent basal images. The basal images generated by ICA are not only irrelevant but also statistical independent, which can characterize the local feature of faces. Face images can be recognized by classifying the coefficients based on SVM (Support vector machines) and CVM (Core vector machines). Both SVM and CVM have high recognition accuracy. With the increasing of ICA feature numbers, CVM can get higher accuracy, less training time and fewer support vectors. Experimental results show that the algorithm based on ICA & CVM is feasible and effective for face recognition.