Emotion Recognition in Video Clips Using Simple and Extended Center Symmetric LBP

The purpose of our visual emotion recognition system is to accurately discriminate six universal emotional states namely anger, disgust, fear, happiness, sadness, surprise. Emotion recognition in humans using visual information has proven to be quite challenging and an active area of research. To the best of our knowledge, simple and extended center symmetric local binary patterns (LBP) have not been used in literature on an emotional data-set and hence we have used it in our work to get a discriminating set of features. Various state of the art classifiers like Support Vector Machines (SVM), Random Forest (RF) and Logistic Regression (LR), are used in our paper to check the performance of our model. A well-known emotional video database known as eNTERFACE’05, has been employed in our research and the highest accuracy obtained from our approach is 97.9%. Our system has yielded higher accuracy on this dataset when compared with other well-known visual emotion recognition systems in literature.