Face recognition using ensemble support vector machine

In Face recognition, a combination of neural network (NN), known as an ensemble of neural network, often outperforms individual ones. This paper is aiming to present a support vector machines (SVM)-ensemble-based efficient face recognition system. The training samples are randomly chosen by means of bootstrap technique to train the different SVM independently. These SVM's are combined together to generate the ensemble SVM. The proposed method then makes a collective decision by aggregating them. It may be noted that, the performance of the practical SVM is far from the theoretical SVM as the implementations are based on approximated algorithms. The performance of the real SVM can be uplifted by using the proposed ensemble SVM with bagging (bootstrap aggregating). Finally, the proposed method takes the collective decision by aggregating the training samples. The proposed method is validated on AT&T, FERET face databases to show its supremacy over the single SVM-based methods.