ISVM for Face Recognition

The similarity of human faces, unpredictable variations and aging are the crucial obstacles in face recognition. To handle this if large set of training images are used then computational complexity will get increase as images are rather high dimension but if training set kept small, performance decreases. Since both classification and feature information are necessary for a recognition system DCT is used to lower the computational complexity and SVM for classification. Since SVM is a popular classification tool but the main disadvantage of SVM is its large memory requirement and computation time to deal with large data set. Therefore we have used incremental learning approach i.e. ISVM to avoid large training time and memory consumption for face recognition. The biggest advantage of using the proposed technique is that it not only decreases the training time and updating time but also improves the classification accuracy rate up to 100%. Experiments are performed on ORL face database and results has proved that not only the training time used by the ISVM is very less compared to SVM but also the recognition rate raised to 100%. Obtained results have presented accurate face recognition system using the proposed approach..