Research Of EEG Emotion Recognition Based On Deep Neural Network

Natural human-computer interaction technology is an important direction of current computer application technology research.Emotion automatic recognition is one of the key technologies to realize natural human-computer interaction.This article focuses on EEG-based emotion recognition.The work done is as follows:An EEG feature extraction method combining empirical mode decomposition(EMD)and approximate entropy(ApEn)is proposed,namely(E-ApEn)for EEG-based emotion recognition.This combined feature extraction method can effectively reduce the complexity of feature extraction,and the extracted feature vectors are sent to an emotion recognition model composed of deep belief network(DBN)and support vector machine(SVM)for training and classification.The experimental results show that the average recognition rate and optimal recognition rate of the four emotions of happy,sad,calm,and fear are 84.34%/90.45%respectively.Based on the previous scheme,this paper proposes an EEG processing method combining Wavelet Transform(WPT)and Hilbert-Huang Transform(HHT)to eliminate the effects of EMD defects.The HHT is used to obtain the EEG's instantaneous energy as a feature vector.Finally,a mixed emotion recognition model composed of a convolutional neural network(CNN),a recursive neural network(RNN),and a support vector machine(SVM)is used for training and classification.Experimental results show that the average recognition rate and optimal recognition rate of the four emotions of happiness,sadness,calmness,and fear are 86.22%/93.45%.Based on the above EEG emotion recognition,this paper fused the EEG and ECG modally to realize the multimodal fusion emotion recognition of EEG and ECG.The instantaneous energy of EEG and ECG is extracted by HHT as their respective eigenvectors,and feature vectors of EEG and ECG are fused into new features through feature fusion and encapsulated into a matrix.Finally,a mixture of CNN,RNN and SVM is adopted to train and classify.The experimental results show that the average recognition rate and the optimal recognition rate of the four emotions of happiness,calmness,sadness and fear are 87.75%/93.85%.