Facial expression recognition using LBP template of facial parts and multilayer neural network

Various methods do exist for facial expression recognition and it's still a challenging task to get accurate results on unseen images due to lack of generalizability. To improve generalization Local Binary Pattern (LBP) are used as feature extraction. Facial parts like eyes, nose and mouth are detected and template is created for each image in dataset. The template of reference image is matched with training image and classified with Neural Network. We train and test with images from the Kohn-Kanade Database. We obtain recognition rate of 95.83% for person independent analysis. Proposed method shows a significant improvement in the recognition rate as compare to existing methods.