A Face Expression Recognition Using CNN & LBP

Visual interaction is an effective means of communication for human beings as social beings. Even a simple change in facial expression signifies happiness, sorrow, surprise and anxiety. The facial expressions of every person should vary in various contexts such as lighting, posture and even background. All these factors still remain an issue while recognizing facial expressions. This paper hopes to bring out a fair comparison between two of the most commonly used face expression recognition [FER] techniques and to shed some light on their precision. The methods being used here are Local binary patterns [LBP] and Convolution neural networks [CNN]. The LBP is meant as a method only for the purpose of extracting features so the Support vector machine [SVM] classifier is being utilized for classifying the extracted features from LBP. The dataset used for the purpose of testing and training in this paper are CK+, JAFFE and YALE FACE. About 70% is being utilized for training and the rest 30% is used in the testing. CNN gives better precision than any other dataset with 97.32% of recognition rate on CK+ dataset and also shows 31.82% of the YALE FACE dataset that is the least accurate we have.