SVMs for few examples-based face recognition

We present an extensive study of the support vector machine (SVM) applied to the few examples-based face recognition problem. The few examples cannot express many conditions that the test data combine, such as changes of pose etc., so we use a simple method to generalize the examples to others. Then, principal component analysis (PCA) is applied to feature extraction from all the images; after that, we use SVM to train and test the data. In the ICT-YCNC face gallery, the proposed system obtains competitive results: a correct recognition rate of 91.59% for all 350 persons.