Autonomic Nervous Pattern Analysis Of People With Trait Anxiety

Nowadays,the fast-paced social life makes people face various stressors and challenges.In addition to physiological stress,psychological stress is gradually eroding people’s mental world.The high trait anxiety populations are more likely to regard external stress stimulus as threatening information,and such cognition makes the high trait anxiety people more anxious.Long-lasted anxiety status will affect social behavior mode of the individuals and even lead to sub-health or disease symptoms related to anxiety.For the health risk caused by high trait anxiety,early intervention benefits from early warning.In order to realize early warning of the health risk of high trait anxiety,this paper studies the stress neurophysiological response of people with trait anxiety.The statistical method is used to analyze the differences of autonomic neurophysiological patterns between high and low trait anxiety populations.Besides,the machine learning algorithm is applied to construct physiological recognition models of high and low trait anxiety for wearable proactive health monitoring.It is helpful to find out the stress autonomic physiological characteristics of high trait anxiety populations in time,so that these people will properly understand their anxiety susceptibility,pay attention to the emergence of anxiety state,actively regulate emotions and promote physical and mental health.The main contents and results of this study are as follows:(1)Design and implement stress elicitation and physiological data acquisition.The high trait anxiety(HTA)and low trait anxiety(LTA)groups were selected by the Trait Anxiety Scale score of State/Trait Anxiety Inventory(STAI),and stress was induced by the Trier Social Stress Test(TSST)paradigm.Meanwhile,data samples were collected from the subjects at rest baseline stage,stress event recall stage,stress event reporting stage and stress recovery stage.(2)Construct the database of high/low trait anxiety population.The database contains electrocardiograph data,stress self assessment data and observers’ stress assessment data of 44 HTA(female: 24;male: 20)subjects and 55 LTA(female: 28;male: 27)subjects.According to the subjects’ and observers’ stress assessment data,the ECG data of the subjects at stress event reporting stage was labeled with stress markers.(3)Analyze the autonomic nervous activity patterns of HTA and LTA in four states,i.e.resting baseline,stress event recall,stress event statement and stress recovery.Taking resting baseline mean heart rate as the normalized denominator,the results showed that there was significant difference in the relative mean heart rate at the stress event recall status between HTA and LTA groups.That is to say,the high trait anxiety populations had stronger stress sensitivity to recalled stress events.Furthermore,the two groups showed significant gender differences in the above-mentioned four states.With TSST paradigm,males in both high and low trait anxiety groups had weaker stress autonomic nervous responses than females.(4)Construct the physiological recognition models of high and low trait anxiety.After backward feature selection and the comparison of multiple classifiers,it is found that the Decision Tree and Naive Bayesian classifier obtained better results of high and low trait anxiety recognition than random guess.The recognition rates of high and low trait anxiety were 68.42%(gender insensitive),58.33%(female)and 66.67%(male)on the validation data set which is independent of feature selection and classifier training.For both genders,the key two-dimension feature subsets of high and low trait anxiety classification came from rest baseline state,stress event recall state and stress recovery state.However,the key feature subsets for different genders contained different RR interval features.The following conclusions are drawn from the above results:(1)High and low trait anxiety populations have differentiated stress neurophysiological response patterns.The classification models of high and low trait anxiety can be obtained by using low-dimension physiological features and classical classifiers.(2)When the amount of data samples is small,the high-dimension features inevitably lead to the over fitting of physiological recognition models for known dataset,and the effective way to avoid over fitting is to limit the dimension of feature space.(3)In real application,only three of the four states of the experimental design,i.e.baseline status,stress event recall status and stress recovery status,are necessary for the recognition of high and low trait anxiety.These three states do not need the statement of stress event and the participation of experts,and the data acquisition is feasible in the practice of wearable health monitoring.