Fusion of Ensemble Technique to equilibrize the Prediction of Face Emotions using Deep CNN models for Real time images

Humans have an uncanny ability to detect and recognise emotions, which is being researched for use in computerization. Face emotion prediction remains a difficult field of study in spite of wide applications due to its subject dependency. Innovative method of using the ensemble classifier for the real time face emotion prediction using base classifiers as Deep CNN models is proposed in this paper. The deep learning algorithms are used as level 1 base classifiers. The imbalance dataset CK+ and small dataset JAFFE are enhanced synthetically by image augmentation method. At level 2, a meta classifier that is a fusion of majority and relative voting techniques is utilized to improve the accuracy of individual emotions. The proposed technique’s overall performance is evaluated and validated using randomly selected face emotion images from the internet with improved overall accuracy. The proposed ensemble fusion technique NCL is used for cross validation on FER2013 dataset.