Research On Emotion Recognition Algorithm Framework Based On The Spatial Association Of EEG Leads

With the development of society,people are under pressure from various aspects,and mental health is challenged.In recent years,mental illness has received more and more attention.With the development of artificial intelligence,machine learning has been widely used in the medical field.In brain research,especially in the study of mental illness,machine learning plays an important role,among them,emotional research is also an important topic,and it is of great significance for the study of psychological disorders.Electroencephalography(EEG)is an electrophysiological signal for recording brain activity.It is combined with machine learning to study emotions is a hot research topic today,but after years of development,the research has encountered a certain bottleneck.The bottlenecks are mainly reflected in three aspects,the limitations of features,the complexity of the algorithm framework and its generalization performance.In the past research,the traditional algorithm framework used artificial features and traditional machine learning to identify emotions.With the development of deep learning,the use of deep learning to extract abstract features based on artificial features has become a new trend,but most of the artificial features consume a lot of computing resources and destroy the original spatial information of the brain area.And due to human specificity,the generalization ability of the model is particularly important.In view of the above three problems,this paper designs an emotion recognition algorithm framework based on the spatial association of EEG leads and EEG-based deep network ensemble learning framework,as follows:(1)Designed an emotion recognition algorithm framework based on the spatial association of EEG leads.Using EEG lead correlation coefficient matrix to characterize the original spatial information in the brain region.Using Convolutional Neural Networks(CNN)to simplify feature extraction process,making full use of the spatial characteristics of CNN to extract the abstract brain space characteristics.It is verified that the algorithm framework effectively improves the accuracy of emotion recognition.(2)Designed an EEG-based deep network ensemble learning framework.Integrating ensemble learning to improve the generalization ability of the framework,optimizing the structure of the CNN,and using two different EEG lead correlation coefficient matrices as input of CNN,it obtain a higher recognition rate than single input.The algorithm framework is lighter than the previous framework,and it improves the recognition rate from the perspective of improving generalization performance.Finally,using the DEAP emotion recognition database,through a variety of emotional classification experiments,it is verified that the emotion recognition algorithm framework based on the spatial association of EEG leads using the original spatial information of the brain is effective in emotion recognition,and verified that the EEG-based deep network ensemble learning framework can effectively improve the recognition rate of emotion classification by optimizing the correlation coefficient matrix and Integrating ensemble learning.