Research On Online EEG Emotion Recognition System Based On Fusion Feature And Ensemble Classification

Emotion,as a high generalization of people's subjective knowledge,is mainly a psychophysiological state produced by feeling,behavior and thought.In the field of human-computer interaction,affective computing is a new developing research direction.With the development of non-invasive sensor technology,machine learning and deep learning technology,emotion computing based on electroencephalogram(EEG)signal has been paid attention by many scholars.Compared with the offline emotion recognition system,the online EEG emotion recognition system pays more attention to emotion feedback and increases the participation in human-computer interaction.Therefore,this paper proposes an online emotion recognition system based on EEG.The main contents of this paper are as follows:(1)The method of Emotion Feature Extraction Based on fusion feature is designed and implemented,which improves the representativeness of emotion feature.Compared with the traditional frequency domain feature analysis method,the combination of time-frequency features can not only characterize the energy characteristics of the signal,but also obtain the phase relationship of the signal.Therefore,the time-frequency analysis method is used to extract the features of the EEG signal.In view of the personal difference of EEG signal and the poor representation of single feature,this paper extracts the fusion feature of EEG signal based on power spectrum feature and wavelet energy entropy feature through cascade feature fusion algorithm and principal components analysis(PCA),and reduces the dimension of feature.The experimental results show that the recognition results based on fusion feature are better than those based on single feature for three kinds of emotions(happiness,calm and sadness).(2)Aiming at the problem that the generalization ability of single classification model to EEG signal is insufficient,this paper proposes an ensemble classification algorithm based on soft voting algorithm.Three kinds of weak classifiers,namely KNN classifier,SVM classifier and RVM classifier,are used as the basic classifiers.On the premise of determining the optimal parameters of the basic classifiers,the advantages of the basic classifiers are retained to the greatest extent.On this basis,the recognition performance of the integrated classifier is improved by soft voting algorithm.The results show that the performance of ensemble classifier is better than that of weak classifier,and the accuracy of emotion recognition based on EEG signal is improved.(3)An online emotion recognition system based on fusion feature and ensemble classifier is designed to realize real-time emotion feedback.Emotion is real-time,online emotion recognition system is conducive to analyze the current emotional needs of subjects,and it is of great significance to improve the participation of human-computer interaction.In this paper,three kinds of emotion movie clips are selected as the stimulus source to induce the subjects' emotion,and the consumer Emotiv is used as the signal acquisition device to complete the real-time acquisition of EEG signal.On the basis of fusion algorithm and integrated classifier design,an online EEG emotion classification system based on MATLAB and python platform is designed to complete the online transmission and real-time classification of EEG signals.Experimental results show that the system designed in this paper can effectively complete online emotion recognition.