Face recognition based on sparse coding using support vector machine classifier

In this paper, a system for face detection based on the generalized BOW method is proposed. We have utilized the space pyramid matching (SPM) method to overcome the neglected problem of space order of BOW. In the feature extraction stage, we have used SIFT method which is resistant against local variations. Sparse presentations usually are linearly separable; hence in the proposed system, we have utilized the sparse codding method in the feature learning stage. In the polling stage, we have used maximum polling operation to reach a unified vector from multiple descriptor vectors. Finally, a support vector machine classifier is used to classify face descriptor vectors. Simulation results show high accuracy of classification (ACC=0.9952) and its resistivity against previous methods.