Automated classification of facial expressions using bag of visual words and texture-based features

As facial expression plays undoubtedly a key role in conveying human emotions and feelings, research into how people react to the world and communicate with each other still stands as one of the most scientific challenges to be addressed. Recent research has shown that facial expressions can be a potential medium for various applications. In this research paper, we explore the use of texture-based facial features obtained using the Local Binary Patterns operator. The facial expression signature is constructed via encoding the textural information using the bag of features. Features are trained to robustly distinguish different seven facial emotions including: happiness, anger, disgust, fear, surprise, sadness as well as the neutral case. Based on a gallery dataset containing 76 images, a classification rate of 93.4% is achieved using the Support Vector Machine classifier. The attained results assert that automated classification of facial expression using an appearance-based approach is feasible with an acceptable accuracy.