Research And Realization Of Emotion Recognition From EEG Based On Deep Learning

Emotion plays an important role in interpersonal communication and medical research.Emotion recognition is an essential part of emotion research and an interdisciplinary field of computer science,neuroscience,psychology and cognitive science.Among them,"Affective Computing" is an authoritative technology.Affective computing can use computers to automatically identify,understand and reflect human emotions.For example,computers can recognize emotions through facial expressions,speech,blinking,posture and physiological signals.Most previous studies focused on methods such as speech and facial expressions.But in some cases,people cannot accurately reflect their emotions into their facial expressions.For example,some people deliberately hide their true feelings for some special reasons,and patients who suffer from facial neuritis cannot express affection.In order to solve the above problems,many researchers have proposed emotional recognition methods based on Electrophysiological(EEG)signals,which are more reliable and more real than the traditional methods.However,t he existing EEG emotion recognition technology has many shortcomings,This thesis is devoted to the research of new methods and technologies to solve the problem of low accuracy in EEG emotion recognition.This thesis elaborates on the methods of emotion recognition,summarizes the research ideas of deep learning,analyzes and compares the advantages and disadvantages of existing EEG emotion recognition schemes,and focuses on the following two aspects of EEG emotion recognition of deep learning:1.Based on the EEG emotion recognition method of ensemble convolutional neural network(ECNN),a five-layer convolutional model is designed and built,and then the plurality voting method in ensemble learning is used to obtain the final result.This scheme can automatically mine the correlation between EEG signals and peripheral physiological signals,automatically extract the effective features and give the results.Finally,we divided emotions into four categories.This experiment is conducted on the DEAP data set.Our experimental results show the effectiveness of our method,and the comparison with related methods shows that the scheme has better performance in accuracy.2.In order to further improve the accuracy and stability of EEG emotion recognition,this thesis adopts the global average pooling(GAP)layer to replace the traditional full connection layer.After the convolutional neural network layer,the GAP is carried out to optimize the instability of the loss rate.It demonstrates the feasibility of this scheme by comparing the full connection layer with the global maximum pooling layer on the DEAP dataset.Finally,the thesis designs an electrical emotion recognition based on in-depth study of the prototype system,through the EEG data input,through the model training criterion to get the final emotional categories,test the accuracy and stability of the emotion recognition at the same time,the experimental results show that the proposed eeg emotion recognition method based on the deep study of the feasibility,accuracy and efficiency.