Face Recognition under Occlusion: An efficient Handcrafted Feature & SVM based Approach

Biometric-based recognition systems have been consistent matter of research and abundant contributions can be seen in different biometric areas. The growing artifacts in face biometrics in terms of occlusions have been a challenge and fooled the best face recognition system. The increased occlusion over the face had drastically reduced the chances of accurate detection. The article present an effective handcrafted feature based machine learning approach for face biometric recognition over CelebA dataset. The proposed system is able to extract prominent quality features from the uncovered region of the pose aligned and background cluttered face regions. The system overrule the need for augmentation for the unbalance dataset and performed better for low samples. The classification accuracy over 100 celebrities with distinct number of samples was found to be 98%.