Study On Feature Combination And Channel Optimization Selection Of EEG For Emotion Recognition

Emotion plays an important role in people's daily life.With the rapid development of human-computer interaction in recent years,the ability to correctly identify and analyze human emotions has become an important criterion for intelligence.Compared with other physiological signals,EEG has become the focus of research on emotion recognition because of its high temporal resolution,difficulty in disguise and close connection with brain activity.In the past,researches of emotion recognition based on EEG signals were mostly developed through the analysis of full channel EEG signals.Although good results were achieved,they do not meet the requirements for portability in real-life applications.To address this issue,this paper studies the effects of different EEG feature combinations and channel optimization selections on emotion recognition,and applies it to a national defense research project.The main works are as follows:1.Firstly,we design an emotion-inducing experiment in which the stimulus was picture,and then we collect emotional EEG data.Fifteen features were extracted from the pre-processed EEG signals,and their performances in the three situations of four emotion classifications,two emotions classification on valence and two emotions classification on arousal were calculated and compared by SVM.Combined with the MRCS channel selection algorithm,this paper studied the effect of channel sorting corresponding to different feature combinations on the accuracy of emotion classification and based on this we proposed a personalized optimal channel selection method.2.For the problem of large differences in the channel sorting between subjects due to individual differences,the paper uses weighted addition method,accurate weighted method,and rank weighted method to study channel selection algorithms that do not rely on subjects and select common effective channels.And we also find out the distribution of the brain regions of these channels.Comparing with individualized channel selection,although the accuracy of the emotional classification of the common channels decreased slightly,the average correct rate of all subjects for the four emotions classification by using the first 10 channels still reached about 95%.3.Considering the fact that the actual wear of brain electrodes is less in the wearable environment,this paper studied the emotion classification of the seven EEG channels Fpz?Fp1?Fp2?AF7?AF8?F7 and F8 in the forehead hairless area and explored the effects of 15 features and different combinations of features on emotion recognition.The results show that the combined features have better accuracy and stability in the emotional classification and the 1st and 2st combinations perform best.With the increase of the number of channels,the accuracy of emotional classification increases,and the standard deviation decreases.Therefore,when designing portable EEG equipment,it is necessary to consider both the number of channels and the classification effect.