Research On Influencing Factors Of Emotional Personalization Based On EEG

Emotions plays a significant role in human cognition,and Electroencephalogram(EEG)is a general reflection of the brain,expressed as the response of electrical signals from brain nerve cells in the cerebral cortex or on the surface of the scalp.In the field of emotion recognition,EEG has many advantages,such as resistance to artifacts and objectivity.Due to the difference of the human body and the different environment,many individual factors will have a certain impact on the emotional recognition of EEG.We studies the influence of various personalization factors in EEG emotion recognition,mainly from three aspects: individual differences,environmental differences and differences in other physiological signals.The other physiological signal chosen is the electrocardiogram(ECG).The main work of this paper is as follows:In terms of individual variability,an EEG-based cross-individual transfer emotion recognition method is proposed.Considering that differences in data distribution due to individual differences can degrade the performance of the cross-individual emotion recognition model,the method is based on the idea of transfer learning.Firstly,the source domain data is selected among existing individuals with labels to prevent negative transfer,and then the manifold embedded distribution alignment method is used to map and align the source domain and target domain feature space to complete the emotion classification.The results show that the accuracy of the knowledge transfer model is more advantageous,with an average accuracy of 74% for positive,neutral and negative emotions.In terms of environmental differences,the effect of different environmental states on EEG-based emotion recognition is studied.Considering that different environmental factors can lead to changes in individual emotions,four environmental states were constructed to evaluate the effect of different environmental states on emotion recognition for two environmental factors: the presence or absence of light and the presence or absence of others’ company.The experimental results show that the accuracy of emotion recognition in a single environment is,in most cases,higher than that in an integrated environment where the four environments are combined.The highest emotion recognition rate is found in the environmental state without light and accompanied by people,and the average recognition rate increases by 13.91%compared with the integrated environment;positive and neutral emotions are more likely to be aroused in the environment with someone else accompanying;negative emotions are more likely to be aroused in the environment with light.In terms of using ECG-assisted EEG signals for emotion recognition,a dimensional emotion recognition method based on a combination of EEG and ECG signals is proposed.For the EEG signal,a support vector machine classifier is used for feature classification,and for the ECG signal,a corresponding bidirectional long short-term memory network model for emotion recognition is established,and the classification results of ECG and EEG are merged through evidence theory.Experimental results showed that this bimodal fusion model outperformed the unimodal emotion recognition model.In both Arousal and Valence dimensions,the average accuracy was improved by2.64% and 2.75% respectively compared to the EEG-based emotion recognition model.Compared to the EEG-based emotion recognition model,the accuracy was improved by7.37% and 8.73%.