Kernel-based scale-invariant feature transform and spherical SVM classifier for face recognition

Due to the wide range of commercial and law enforcement applications and the availability of feasible technologies, face recognition has recently gained significant attention among the researchers. The literature presents various face recognition systems, which are capable of measuring and matching the distinctive features intended for the purpose of identifying or verifying a person from a digital image. The identification of distinctive features from the face image poses various challenging aspects due to the various poses and illumination conditions. To overcome these major limitations in the existing methods, this paper proposes kernel-based Scale-Invariant Feature Transform and spherical SVM classifier for face recognition. Furthermore, a novel weightage function for feature extraction and classification, which is termed as Multi Kernel Function (MKF), is also proposed. To extract facial features, we adopt SIFT technique, which is modified in the descriptor stage by the proposed MKF weightage function, thereby evolving a new technique which we termed as KSIFT. Multi-kernel Spherical SVM classifier is used for the classification purpose. The performance of the proposed method is analyzed by performing experimentation on CVL Face Database for the evaluation metrics, such as FAR, FRR, and Accuracy. Then, the performance is compared with the existing systems, like HOG, SIFT, and WHOG. From the experimental results, it can be shown that the proposed method attains the higher accuracy of 99% for the face recognition system.