Research On Emotion Classification Based On EEG Signal

The brain is the most complex part of human organ which contains rich body’s physiological and psychological information. Changes of Physiological and psychological will be reflected in the brain. Currently the field of brain research has become an important area of research. Among them, emotional and brain research is becoming a hotspot research that researchers want to study. Many researchers at home and abroad using EEG machine classify the different emotions. They aim to enhance classification accuracy of emotions and to explore different emotional adjustment mechanism. The study of emotions not only helps treat depression disease but also has a very important significance in BCI applications.Draw on the experience of existing brain research results, this topic processed and analyzed different emotional levels of eight collage subjects’ EEG signals which collected under high-valence、low-valence、high-arouse、low-arouse. After we get original signals, we filter and denoise the EEG signals. Then we use Wavelet Transform method to get δ,θ, α,β four EEG band which bands energy are as eigenvalues from32leads. According to the experimental program of this study, we choose the improved SVM classifier to classify high-valence、low-valence、 high-arouse、low-arouse samples which come from the group and the individual in order to find if there is a significant difference between two kinds samples. At last, we analysis two kinds features which is selected by Relief algorithm. From that we want to find most relevant brain regions and EEG frequency. At last, the research use Relief algorithm to select features, aiming to find the relevance brain regions and bands and try to find reasonable explanations.The study found that groups’ EEG signals on the valence and arousal are inseparable, which means there are no differences of different emotions and there is not the same pattern or regulation. But it is obviously that individuals’ EEG signals are separable for different emotions. Brain regions associated with high-valence、 low-valence mainly exist frontal lobe. Alpha rhythm of right frontal lobe is active under high-valence emotional state. Low-valence emotional state evokes left frontal lobe, and right frontal alpha rhythm wave shows inhibition. Brain regions associated with high arousal-low arousal is located in the frontal and temporal lobes. High arousal of emotional states mainly related to the temporal lobe, and low arousal of emotional states mainly associated with frontal and temporal lobe. For individuals, energy of alpha, beta under low arouse state is higher than that under high arouse state, beta waves is active when she or he is excited. Alpha, beta, theta frequency under high-low arouse of the brain in the temporal lobe are quite familiar.The process of research is drove by data. Draw lessons from the existing research achievements of neuroscience and machine learning methods, so that the result of research is more reliable. The research improves the classification performance of SVM classifier, which is helpful for the application of emotion recognition. Meantime, the research study relevance of brain regions under different emotions, which reveals the different brain mechanism of emotion regulation and provides ideas and methods for the research on emotional brain mechanism.