Face Recognition Based On Support Vector Machine And Edge Information

The technology of human face recognition has been one of the popular problems in Pattern Recognition and computer vision. Now there are many approaches on face recognition, such as eigenface, Neural Network approach, Hidden Markov Model (HMM) and face recognition based on geometry feature. But the importance of edge information was ignored. So the using of edge information is discussed in this paper and a robust classified machine based on the support vector machine (SVM) is built.After comparing some approaches of feature abstract and recognition, a novel approaches suggested. In this method face image and its edge information are combined to enhance features. Then, the discrete cosine transform (DCT) is use to abstract feature vector. Last, SVM is used to classify these feature vectors. Edge information in face image not only can describe face features with less information, but also is not sensitive to the change of light. This method can emphasize features in face and make full use of the information. The discrete cosine transform is one of the techniques of image compress. And it can also be used to abstract features. DCT can concentrate energy on lower domain. So when the number of feature vectors is reduced, the distribution of feature sub-space can be retained. The support vector machine is a new statistical learning method. It can solve small-sample, non-linear and high dimension problems by using structural risk minimization (SRM) instead of empirical risk minimization (ERM).Now it is applied in Pattern Recognition, regression estimate and probability estimate.This approach was tested on ORL face databases and the results demonstrate that it performs very well when compared with other approaches.