Publication

Abstract : The use of the Electroencephalogram (EEG) has been highly promising in regards to emotion recognition. Emotion detection is crucial in the areas of treatment and medical care. Emotion recognition in humanmachine interaction allows for an easier process of identifying emotions. Recently, advancement in technology has made the treatment of emotionally-ill patients more convenient because doctors can now provide more appropriate medical care to them. All around, real-time assessment and management of emotion improves the quality of human life. The main objective of the research conducted is to analyze human being’s different emotions: fear, sadness, happiness and neutral and assess the accuracy of machine learning methods. We have investigated the SEED-IV dataset for computing the emotions and various models have been employed to classify them. The dataset acquired was pre-processed and labeled with, accordingly. To add, we have used the Differential Entropy (DE) feature on the gamma frequency range from 62 different channels of the EEG to train the models. The classifiers used were KNN, Random Forest, SVM, and Neural network. The accuracies obtained were 70%, 83%, 85%, 80% from KNN, Random Forest, SVM, Neural network, respectively.