A Study Of Emotional Tendency Based On Deep Learning And EEG Signals

Human emotions,including people's psychological reactions to external or self-stimulation and the physiological reactions accompanying such psychological reactions,often play a very important role in interpersonal and decision-making processes.Computer-based automatic emotion recognition,because of its natural objectivity and powerful generalization ability based on big data,is of great significance for building more intelligent human-computer interaction,disease diagnosis,psychological counseling and other systems.The dissertation mainly studies the emotional classification based on EEG signals and the mechanism of emotional activation,focusing on the current hot and cutting-edge deep learning techniques.In view of the problems of poor generalization of the model,unclear emotional activation mechanism and lack of psychological theory,two aspects of work were carried out,namely emotional classification and emotional activation mechanisms.Compared with traditional research,which only focuses on the accuracy of emotional classification,this dissertation uses the techniques of deep learning to improve the recognition accuracy,and innovatively studies the emotional activation mechanism,and confirms the results with psychological theory and practice results.The conclusion of this dissertation is strongly supported by psychology,which provides a new way of thinking for the analysis of emotional tendency based on EEG.The main work and innovations of this dissertation are as follow:1.Propose an emotion recognition algorithm based on deep learning.The main idea of this method is to reduce the information redundancy of the input data by extracting the relevant features of the EEG signal,and select a small number of features instead of the high-dimensional EEG signal input into the deep network,and use the deep learning technology to automatically extract the higher-order features with stronger discriminative power.The emotional states are then classified based on this higher-order feature.At the same time,in response to the unreasonable set of experiments in the traditional method,a more reasonable experimental setting method is adopted to improve the generalization performance of the algorithm.Finally,for the emotion classification task on the DEAP dataset,the accuracy score is 87.27%,indicating the effectiveness of the proposed method.2.Aiming at the fact that traditional research only pays attention to the accuracy of emotion recognition and there is less ralated research on emotional activation mechanism,a research method of emotional activation mechanism based on machine learning is proposed.The main idea of this method is to construct an emotional activation curve using the classification results,intuitively reflects the activation process of emotions.After obtaining the activation curve,it is mutually confirmed with the theory and practice of psychology,which provides strong support for the conclusions obtained in this dissertation.The two works in this dissertation cover the application and theoretical research of emotional tendency analysis.While the accuracy of emotional classification exceeds traditional research,this dissertation breaks the gap in this research field by studying the mechanism of emotional activation based on EEG signals.At the same time,combined with psychology,it provides a new way of thinking for the analysis of emotional tendency based on EEG signals.