Emotion Recognition Based On EEG Signals

In recent years,with the progress of science and technology and the rapid development of artificial intelligence,there have been more and more researches on the emotion recognition of EEG signals in physiological signals generated by the brain.The EEG signal is a kind of physiological signal generated by people's central nervous system.Human emotion is closely related to the central nervous system.The brain produces different emotions under different emotional state.We can get corresponding emotional state from different EEG signals,and this physiological signal is not vulnerable to a person's subjective control.So people's emotional state can be reflected by the EEG signals real-time and truly.In the study of EEG emotion recognition,how to extract useful features from EEG signals and choose a good method for classification are two important tasks.The effective emotional features are very important to the emotion recognition.After extracting the effective emotional features,choosing a good classification algorithm for emotion recognition also plays a very important role.This paper mainly studies the emotion recognition from two aspects of feature extraction and classification method.The main work is as follows:(1)The EEG signals were filtered to four frequency bands of Theta(4-7 Hz),Alpha(8-13 Hz),Beta(14-30 Hz)and Gamma(31-45 Hz),and we calculated the corresponding Power Spectral Density on each frequency band.Firstly,Light GBM was used to prioritize the Power Spectral Density features and screen out the effective features EEG emotion classification.Then we also used Light GBM to do the emotion classification.At the same time,SVM algorithm was used for classification and comparison research.The results showed that the classification effect of SVM on extracted features was not as good as that of Light GBM.Moreover,after feature extraction,the accuracy and efficiency of EEG classification can be improved by using the extracted features which are highly correlated with emotion.(2)Based on the Power Spectral Density in frequency domain features,a new classification method of Convolutional Neural Network was proposed to classify EEG signals.This method can effectively classify EEG signals based on Power Spectral Density.At the same time,Random Forest,KNN and Decision Tree were also used to classify the Power Spectral Density features.Firstly,based on the classification method of Convolutional Neural Network proposed in this paper,experiments on binary classification and three classification for Valence and Arousal were did respectively.Then binary classification experiments for Valence and Arousal were did using Random Forest,KNN,and Decision Tree.We compared the binary classification results obtained by using different classification methods.The results showed that the accuracy of binary classification and multi-classification for Valence and Arousal based on Power Spectral Density features had been enhanced.