Identification of Human Emotions from FEG Signals Using Deep Learning Model

Abstract

Emotion is a condition that completely presents human thoughts, feelings, and attitudes, every human carries emotion. Human emotions are being detected through electroencephalogram signals. Here the targeted data set is DEAP data set. Previously a lot of work has been done on human emotional state detection. There are multiple wAYs to detect emotions which include facial expression recognition, verbal recognition, and recognition through electroencephalogram signals. Huge amount of existing work is presently related to the identification of the human emotional state. Including the most recent work done which is in 2020 by Vikrant Doma & Matin Pirouz, who worked in same domain. They used DEAP dataset and algorithms used are svm, knn and decision tree. And they concluded knn is best among all the machine learning algorithms they used. Accuracy of each is ranging between 55-75%. Previously scholars mainly focused on machine learning algorithms for the detection of human emotional state but from 2017 they have rapidly started to use deep learning algorithms for the detection. Rapid focus on deep learning is due to its popularity for dealing huge amount of data very smartly and best accuracy results. So, here deep learning algorithm (CNN) is being focused for the detection of human emotional state. Deep data set is being used which undergoes through the process of signal refining, and then DWT is being used to extract different frequency bands. As a next step mean (mean of detailed coefficient cd) is being computed. After that 12 emotions are being classified through CNN algorithm. At last, 97% accuracy is obtained in this work.