The Study Of Characteristics Of Spectral Asymmetry In EEG Signals And Emotion Recognition

Researchers have been studying emotion recognition for a long time.The brain is the most advanced part of human nervous system.When people are subjected to external stimulation,their brains will produce bioelectrical signals.Electroencephalogram(EEG),EEG contains abundant physiological and psychological information.Researchers pay more and more attention on emotion recognition research based on EEG.With the development and innovation of device technology,using non-implantable electrodes to collect EEG signals is more convenient and faster,and the related research on EEG signals have been booming.Compared with the traditional emotion recognition research methods,emotion recognition research based on EEG signals has many advantages,such as less camouflage and higher accuracy of emotion recognition.With the development of artificial intelligence technology,related research based on EEG signals has also provided a good method for human-computer interaction.At the same time,the research of emotion recognition based EEG signals also provided an effective method for the study of mental diseases.EEG signal is a kind of sequence bioelectrical signals,which contains abundant time-domain features,frequency-domain features and timefrequency features.We can study the brain mechanism of emotions and use key features of people to recognize emotions features through the study of relevant features.This paper used the existing emotion recognition research results based on EEG,Using the images of the international emotional picture system to stimulate the positive,neutral and negative emotions and collect the EEG signals,The collected EEG signals are processed.Using the independent component analysis method to preprocess the original EEG data to remove various noises and artifacts to obtain pure EEG data.The previous research in our laboratory showed that the accuracy of emotion recognition is higher when we use the positive potential as the characteristic in EEG signals to recognize emotion.According to the characteristics of late positive potential,The data of EEG signals from different brain regions and electrodes(segment 1000-1500ms)were processed by choosing their late positive potential segment,and the average EEG signals' spectral asymmetry index(SASI)was calculated when subjects were in positive emotion,neutral emotion and negative emotion.The results showed that,compared to neutral emotion,the EEG signals' spectral asymmetry index was significantly larger when subjects were in negative emotion;Compared to neutral mood,the EEG signal spectral asymmetric index was significantly smaller when subjects were in positive emotion.The SASI had strong characteristics feature especially in left temporal,right temporal and occipital.It is suggested that SASI can be used as one of the characteristic features of emotions recognition.Inspired by the result,this paper proposes a research method combining the late positive potential of the time domain feature with the spectral asymmetry index of the frequency domain feature in the segment of the EEG signals to recognize the emotion kinds.According to the characteristics of the late positive potential to select the EEG data segment.The spectral asymmetry index features of different brain area were extracted in the corresponding EEG data segment when subjects were in three emotional states,and the KNN algorithm was used to explore the emotional classification and results.The classification results showed that the average accuracy of the emotion recognition in the left and right temporal regions was more than 80% when subject were in three emotional states.In addition,the average accuracy of the occipital lobe region when subject were in three emotional states was also more than 75%,The emotion recognition had a good result in the temporal,right and occipital regions,and these three brain regions can be used as the key brain regions for emotion recognition.It shows that the spectral asymmetry index in EEG signals has strong characteristics feature and can be used as one of the important features for emotion recognition.