脉搏传递时间序列用于评估社交焦虑的研究

发布时间：2018-10-26 09:37

【摘要】：社交焦虑属于一种负性情感,普遍存在于大众人群之中,会对人的社会功能和心血管系统造成损害。此外,社交焦虑通常还有可能发展成为社交焦虑障碍,给人们身体和精神上带来极大的痛苦和折磨。因此,及时评估人们的社交焦虑,从而采用有效的情感调节方案来帮助他们摆脱社交焦虑是极为有意义的。社交焦虑会导致血压的升高,而脉搏传递时间与血压高度相关,可作为反映血压的指标。因此,本文对脉搏传递时间序列用于社交焦虑评估的方法进行了探究。本文通过有观众和无观众即兴演讲实验来分别诱发被试的高度焦虑和低度焦虑。根据平均观众评分得到了有观众演讲阶段的30个焦虑被明显唤起的被试。经过T检验发现他们在有观众演讲阶段的相对焦虑程度显著地大于在无观众演讲阶段的相对焦虑程度。脉搏传递时间序列用于社交焦虑评估的实现步骤如下:首先,对脉搏传递时间序列进行去除基线差异和去除极端数据的预处理;而后,提取了统计特征、非线性特征、基于小波包分解的能量特征和矩特征;之后,采用后向选择(SBS)算法进行特征选择;最后,将极限学习机和留一被试法交叉验证用于低度焦虑和高度焦虑的识别与分类。本文经过研究主要有如下发现:(1)基于脉搏传递时间序列,进行去除基线差异和去除极端数据的预处理后,提取的一些统计特征、非线性特征、基于小波包分解的能量特征和矩特征对低度焦虑和高度焦虑有很好的区分。根据SBS算法进行特征选择得到了0阶和19阶加权Krawtchouk矩的特征组合,并用于低度焦虑和高度焦虑的识别与分类,其中对于低度焦虑和高度焦虑的识别率均可达到96.67%。(2)通过对脉搏传递时间序列极端数据去除前后进行对比分析发现:去除脉搏传递时间序列中的极端数据可提高社交焦虑评估的准确性,对去除极端数据前的脉搏传递时间序列提取0阶和19阶加权Krawtchouk矩用于低度焦虑和高度焦虑的识别与分类,其中对于低度焦虑和高度焦虑的识别率仅分别为90%和86.67%。(3)通过对基于脉搏传递时间序列与基于心电RR间期序列的两种社交焦虑评估方法进行对比分析发现:对心电RR间期序列经过去除基线差异和极端数据的预处理后,提取0阶和19阶加权Krawtchouk矩用于低度焦虑和高度焦虑的识别与分类,其中对于低度焦虑和高度焦虑的识别率仅分别为73.33%和70%。本文经过深入研究,最终得出结论:脉搏传递时间序列可用于实现社交焦虑的准确评估。
[Abstract]:Social anxiety is a kind of negative emotion, which is common in the public, and can damage the social function and cardiovascular system. In addition, social anxiety can also develop into social anxiety disorders, causing great physical and mental pain and suffering. Therefore, it is very meaningful to evaluate people's social anxiety in time and adopt effective affective adjustment programs to help them get rid of social anxiety. Social anxiety can lead to higher blood pressure, and pulse transmission time is highly correlated with blood pressure, which can be used as an indicator of blood pressure. Therefore, this paper explores the application of pulse transmission time series to the assessment of social anxiety. In this paper, the high anxiety and low anxiety of the subjects were induced by the experiment of improvisation speech with and without audience. According to the average audience score, 30 subjects with significant anxiety during the speech stage were found to be significantly aroused. It is found that the degree of relative anxiety in the stage of speech with audience is significantly higher than that in stage without audience through T test. The implementation steps of pulse transmission time series used for social anxiety assessment are as follows: firstly, the pulse transmission time series is preprocessed to remove baseline differences and remove extreme data; Then, the statistical feature, nonlinear feature, energy feature and moment feature based on wavelet packet decomposition are extracted. Then, the backward selection (SBS) algorithm is used for feature selection. Finally, the extreme learning machine and the reserved one test method are used to identify and classify low anxiety and high anxiety. The main findings of this paper are as follows: (1) based on pulse transmission time series, some statistical features and nonlinear features are extracted after removing baseline differences and removing extreme data. Based on the energy and moment features of wavelet packet decomposition, there is a good distinction between low anxiety and high anxiety. According to the SBS algorithm, the feature combinations of 0-and 19-order weighted Krawtchouk moments are obtained, which are used to identify and classify low anxiety and high anxiety. Among them, the recognition rate of low anxiety and high anxiety can reach 96.67. (2) by comparing and analyzing the extreme data of pulse transmission time series before and after removal, it is found that the pole of pulse transmission time series is removed. End data can improve the accuracy of social anxiety assessment, The 0 and 19 order weighted Krawtchouk moments were extracted from the pulse transmission time series before the extreme data were removed for the recognition and classification of low and high anxiety. Among them, the recognition rates of low anxiety and high anxiety were only 90% and 86.67% respectively. (3) two social anxiety evaluation methods based on pulse transmission time series and ECG RR interval sequence were compared. It was found that the ECG RR interval sequence was preprocessed by removing baseline differences and extreme data. The 0-order and 19-order weighted Krawtchouk moments were extracted for low and high anxiety recognition and classification. The recognition rates for low anxiety and high anxiety were only 73.33% and 70%, respectively. In this paper, we conclude that pulse transmission time series can be used to accurately assess social anxiety.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R749.72

【相似文献】