Evaluation Of Social Anxiety Through Analysis Of Pulse Transit Time Series

Social anxiety(SA) is a kind of negative emotion which is highly prevalent in the general population, usually causing the impairment of social function and cardiovascular function of people. What’s more, SA usually would worsen to be social anxiety disorder, bringing mental and physical tortures to people, and timely evaluation of SA can help people stop their anxiety through personally effective ways of emotion regulation. Since SA would lead to the elevation of arterial blood pressure(ABP), and the pulse transit time(PTT) has a high correlation with ABP and can be used as an indicator of ABP, this paper puts forward to use the PTT series to evaluate SA and explores the effectiveness of it.In this paper, the subjects’ high anxiety state and low anxiety state are respectively induced by impromptu speaking with 21 audience and without audience. According to the average audience score, it could be determined that 30 subjects have shown apparent anxious symptoms during impromptu speaking with 21 audience. For these 30 anxious subjects, the t-test result shows that the relative anxiety degree during impromptu speaking with 21 audience are significantly higher than that without audience. The implementation steps of the PTT series used for the evaluation of SA are as follows: Firstly, for the preprocessing of the PTT series, getting rid of the influence of the different baseline and discarding the extreme data are necessary. Secondly, for the feature extraction, statistical features, nonlinear features, energy features that based on wavelet packet decomposition and moment features are extracted. Thirdly, for the feature selection, sequential backward selection(SBS) algorithm is used. Finally, extreme learning machine(ELM) and leave-one-subject-out cross validation are used for the recognition and classification of low anxiety state and high anxiety state.The main conclusions of this paper are as follows:(1) Some extracted features such as statistical features, nonlinear features, energy features that based on wavelet packet decomposition and moment features of the preprocessed PTT series are good features for distinguishing low anxiety state from high anxiety state. According to the SBS algorithm, the 0 order Krawtchouk moments and the 19 order Krawtchouk moments are selected as the best features for the recognition and classification of SA. And with the combination of these two features used for the recognition and classification of SA, it turns out to be that the recognition accuracy of low anxiety state and high anxiety state are both 96.67%.(2) Through the comparative analysis of the PTT series that with extreme data and that without extreme data for the evaluation of SA, it turns out to be that with discarding the extreme data, the recognition accuracy of SA could be improved. For the PTT series that with extreme data, this paper extracts the 0 order Krawtchouk moments and the 19 order Krawtchouk moments as features used for the recognition and classification of SA, and the result shows that the recognition accuracy of low anxiety state and high anxiety state are 90% and 86.67%, respectively.(3) Through the comparative analysis of the PTT series and the RR interval series that used for the evaluation of SA, it could be found that when the 0 order Krawtchouk moments and the 19 order Krawtchouk moments of the preprocessing RR interval series are extracted as features used for the recognition and classification of SA, and the result shows that the recognition accuracy of low anxiety state and high anxiety state are 73.33% and 70%, respectively.Through the intensive study, this paper finally concludes that the accurate evaluation of SA could be well achieved through the PTT series.