Facial expression recognition and emotion classification system for sentiment analysis

A huge amount of data is available to the web users with evolution of web technology. The available resources in web are used by the users and also they involve in giving feedbacks and thus generate additional information. It is very essential to explore, analyze and organize their opinions and feedbacks in an efficient way for better decision making. So sentiment analysis (SA) is used to find the users opinion about a particular topic, product or problem in a more efficient way. The main aim of SA is to solve the problems in relation to opinions about products, movies, politics, review sites etc. Sentiment analysis can be done with different modalities by taking inputs from text, image, audio and video. This paper proposes a system that will automatically recognise the facial expression from the image and classify emotions for final decision. The system uses a simplified method called `Viola Jones Face Detection' algorithm for face localization. Then facial features are extracted using three methods `Zernike moments', `LBP' and `DCT transform'. The different feature vectors are combined together using a subset feature selection algorithm to improve the performance of recognition and classification process. Finally the combined features are trained and classified using SVM, Random Forest and KNN classifier. Experiments are conducted on JAFFE database and finally performance of the system is evaluated.