Face Recognition under Partial Occlusion using HMM and Face Edge Length Model

There are many applications which use Face Recognition for identification or verification of a person. In this study, a face recognition system based on HMM has been proposed to handle the problem of partial occlusion. Face is represented by eight isolated regions: Hairs, Forehead, Eyebrows, Eyes, Nose, Upper Lips, Mouth and Chin. The non-occluded region in face image of testing and training image is used for processing. Further, to increase the accuracy and robustness in the face recognition system HMM is coupled with Face Edge Length Model (FELM) in recognition phase. FELM contains various lengths between the any two edge points on the face. The proposed model is more flexible as it handles general occlusion. Experiments are performed only for sunglasses and scarf occlusions in AR database. Experimental results reveal that the proposed algorithm outperforms state-of-art as well as those methods that uses only HMM in recognition phase.