Emotion Recognition Research Based On EEG Signal

This paper mainly study the emotion recognition based on EEG signal, according to the characteristics of EEG signal and the formation mechanism of emotion, based on the existing related research, emotional classification of EEG signal is studied in the frontal lobe brain areas. With the DEAP database analysis of the experiment, Based on the valence and arousal in the two-dimensional emotional dimension model, the classification standard of pressure state and calm state is selected as the emotional category. The main point of the paper is as follows:(1) Researching the EEG feature extraction methods. Using linear analysis of wavelet transform, wavelet energy is extracted from EEG under different rhythms; nonlinear dynamics methods to extract the nonlinear characteristics of EEG emotions are used, including approximate entropy, sample entropy and permutation entropy.(2) Studying the EEG pattern classification methods. Support vector machine(SVM) classification method is mainly used to study the parameter selection of SVM punishment factor and RBF kernel function, then the parameters are optimizated through improved grid search method and verified by experiment.(3) Researching the emotion recognition based on EEG by experiment. Using wavelet energy and three nonlinear parameters, and analysising of EEG signals by pressure state and calm state, recognition and classification were studied from a single characteristic parameter and fusion of multiple characteristic features.The results show that: in the view of feature parameters, classification effect of nonlinear analysis method is better than that of linear analysis method, especially, permutation entropy gets the best classification effect. in the view of dimension analysis, classification effect of fusion multiple characteristic parameters is better than that of single feature parameter, the obtained classification accuracy of the fusion of three nonlinear feature parameters reaches 77.8%.