Fusion Emotion Prediction Using the CEFER Algorithm

Recent research on facial expression identification with a single tag typically covers the seven main emotions like happiness, sadness, fear, disgust, anger, and surprise as well as the neural facial expression. In recent days, compound emotions are getting more attention since they are essential for numerous biological and social applications. When an external stimulus is formed, depending on environmental circumstances, a person will produce compound emotions. The previous studies have identified 37 more complex emotions in addition to those basic emotions. However, they proved incapable of evaluating certain other emotions, and a few were eliminated by classifying them as meaningless. In this research, a novel CEFER (Compound Emotion Facial Expression Recognition) algorithm was developed to extract compound emotions and classify the different emotion types. In the design presented here, a combination of 50 emotions is recognised as mixed emotions, 13 of which are newly developed. This work focuses on the recognition of basic and compound emotions using a number of databases, including CFEE (Compound Facial Expressions of Emotion), naturally occurring affective faces (RAF) etc. A deep feature extraction model was created using transfer learning based on residual neural models, and several machine learning techniques were subsequently employed for classification. In addition, various SVM kernels, such as linear, polynomial, and sigmoid, as well as random forest and logistic regression methods, are trained for result comparisons. In comparison to SVM kernels the fundamental emotion and compound emotion average accuracy achieved 98.78%, 82.36%, 94.35% and 75.35% by SVM-RBF for both above mentioned databases. This research seeks to identify seven fundamental and forty-three complex facial emotions.