Physiological Network Of Emotion

Emotion is a physiological and psychological process,which plays an important role in people’s lives.Emotional experience is highly related to the responses of the physiological system,while Physiological responses constitute the basis of emotional experience.Emotional perception changes the states of the physiological systems,so that people can better adapt to unfamiliar conditions and challenges.However,the previous research on emotions has almost focused on how emotions change the activities of individual physiological systems.Recent studies have shown that human physiological systems are a complex whole that continuously interacts and forms a dynamic network.Therefore,this article focuses on the research of affective analysis based on physiological networks.The main research includes the following three aspects:1)This article adopted that the multivariate transfer entropy is used to estimate the physiological information interaction between different physiological systems.The transfer entropy quantifies how one system affects another system from the perspective of information theory,and has been proven to be able to accurately evaluate all types of interaction between physiological systems.Then,six kinds of physiological signals,that are highly related to emotions,including electroencephalogram(EEG),electrical skin response,photoplethysmogram signal,electromyographic signal,respiration signal and body temperature,are used to construct physiological networks across forty emotions.Different from the traditional brain network model based only on EEG signals or the individual physiological signal characteristics only focus on the physiological interactions within a single system,the physiological networks consider the dynamic interaction among all these physiological systems,and explores the physiological changes under different emotions from a holistic perspective..2)This paper developed five types of physiological networks under different emo-tions and different EEG signal frequency bands,and provided physiological interaction networks under different EEG signal frequency bands.We found that under different emotions,the physiological network topology is significantly different.By calculating the global topological parameters of these physiological networks,we found that the physiological network based on the original EEG has a significant correlation with emotional valence,and the physiological network based on theta band has a significant correlation with the emotional arousal.We also found that the heart system and respiratory system play an important role in the physiological network,and found that different physiological systems in the physiological network have different sensitivity to emotional valence and arousal.By taking the peripheral physiological systems as a whole,we also propose central-peripheral physiological network.We proved the important role of peripheral ner-vous system(PNS)for physiological network.And we found that the interaction from the centeral nervous system(CNS)to PNS is significantly different to the interaction from the PNS to CNS,and the interaction between the CNS and PNS changes relatively with emotions.3)We also apply the physiological network to emotion recognition.We extracted the topological characteristics of the physiological network and the characteristics of traditional single-channel physiological signal as comparison.We performed quantitative emotional predictions on the two types of features through 7 machine learning models.We found that the physiological network topology features achieved better performance in four models,and similar performance in remain three while the number of network features is obviously less than traditional features.It shows that the physiological network method has great potential in the field of affective computing.In sum,we proposed a system-wide integrative approach,referred to as the physiological network,in order to study the physiological responses of emotion.It not only contributes to the deeper understanding of emotional mechanisms but exhibits potential for disease research and engineering applications.