基于HRV信号的压力识别及特征对比分析

发布时间：2018-02-03 06:34

本文关键词： 心理压力心率变异性序列后向选择人工神经网络支持向量机　出处：《西南大学》2017年硕士论文　论文类型：学位论文

【摘要】：现代社会中,越来越多的人受到心理压力的困扰,不同程度的心理压力会对人产生生理和心理上的影响。此外,长时间的心理压力还可能引发抑郁,从身体和心理上给人带来极大的痛苦和折磨。因此,如何能够准确评估人们所处的压力程度,从而采取有效的调节手段是近几年研究的热点。目前对于压力的评估主要分为基于心理问卷的压力评估和基于生理参数的压力评估两大类,前者虽然更容易量化且便于统计分析,但是往往需要参与者积极的响应和配合,更容易受到个体主观的影响,这种方法得到的压力识别结果往往被认为是不可靠的。当人体受到心理压力刺激时,人体的生理平衡状态将会发生变化,这会直接导致人体各项生理参数的改变。因此,本文提出采用心率变异性信号对于人体所处的压力状态进行识别和评估,并在此基础上,对于能够反映心脏自主神经活动的心率变异性指标进行分析,寻找其随着压力程度的增大而表现出的变化规律,从而可方便的实现对心理压力程度的监测,更好的实现心理压力的识别与评估。本文的主要研究内容如下:(1)参考心算任务范式,设计诱发心理压力的实验方案,并对实验得到的量表和任务绩效两项指标进行分析,将实验诱发的心理压力分为三个程度。(2)对收集到的原始ECG信号进行预处理。在本文中,采用小波变换的方法去除了ECG信号的噪声干扰。对滤除噪声后的信号经过R波定位,提取出HRV信号。从时域、频域以及非线性三个方面提取HRV特征共计26维,并在此基础上对特征进行预处理。(3)建立心理压力程度三分类模型。采用SBS算法,构建符合实验要求的BP网络模型和SVM分类器,从提取的原始特征中选择出对分类贡献度最高的特征集,并对两种分类器的分类准确率通过留一验证法进行评估发现两种分类器的识别率均在80%以上。(4)通过重复测量方差分析对第四章选取的不依赖于本文构建的两种分类器的特征集中的每个特征进行分析,判断在不同心理压力程度下每个特征的显著性差异水平。找出随着压力程度的变化,特征的变化规律,并对其造成的神经活动机制进行分析。本文经过深入研究发现,参考心算任务的实验范式,通过时间来控制心算任务难度能够诱发出不同程度的心理压力,采用SBS算法,并构建符合实验要求的BP网络模型和SVM两种分类器,将SBS与两种分类器分别结合得到的两种分类器的平均识别率均在80%以上,并选择出对两种分类器贡献度最高的特征组合。说明心脏自主神经活动可以在一定程度上区分心理压力。随着压力程度的增大,HRV指标中的SDNN显著减低而aHF显著升高,由此可以推测出随着压力程度的升高,心脏交感-迷走神经活动产生了改变,交感神经功能在一定程度上受到抑制而迷走神经功能则相对兴奋。我们可以参考这几项指标的变化规律对压力程度进行检测和评估。
[Abstract]:In modern society, more and more people are troubled by psychological pressure, different degrees of psychological pressure will have a physiological and psychological impact on people. In addition, long-term psychological stress may also lead to depression. Physical and psychological pain and suffering. Therefore, how to accurately assess the degree of stress people are in. In recent years, stress assessment is mainly divided into two categories: psychological questionnaire based stress assessment and physiological parameter based stress assessment. Although the former is easier to quantify and easy to analyze, it often needs the active response and cooperation of the participants and is more susceptible to the subjective influence of the individual. The results of this method are often regarded as unreliable. When the body is stimulated by psychological stress, the physiological balance of the body will change. This will directly lead to the changes of human physiological parameters. Therefore, this paper proposes the use of heart rate variability signal to identify and evaluate the pressure state of the human body, and on this basis. The indexes of heart rate variability which can reflect cardiac autonomic nervous activity are analyzed to find out the change law of heart rate variability with the increase of stress degree, so that the monitoring of psychological stress degree can be realized conveniently. The main contents of this paper are as follows: 1) referring to the task paradigm of mental arithmetic, the experimental scheme of inducing psychological stress is designed. The scale and task performance were analyzed and the psychological stress induced by the experiment was divided into three levels. (2) the original ECG signal was preprocessed. In this paper. The noise interference of ECG signal is removed by wavelet transform, and the HRV signal is extracted from time domain by R-wave localization of filtered noise signal. On the basis of extracting 26 dimensional HRV features from frequency domain and nonlinear three aspects and preprocessing the features, a three classification model of degree of psychological stress is established. SBS algorithm is adopted. The BP neural network model and the SVM classifier which meet the requirements of the experiment are constructed, and the feature sets with the highest contribution to the classification are selected from the original features extracted. The classification accuracy of the two classifiers was evaluated by a residual verification method. It was found that the recognition rates of the two classifiers were above 80%. Through repeated measurement variance analysis, we analyze each feature in the feature set of the two classifiers which are not dependent on the two classifiers constructed in this paper in Chapter 4th. Judge the significant difference level of each characteristic under different degree of psychological stress. Find out the change rule of characteristic with the change of stress degree. After in-depth study, this paper found that referring to the experimental paradigm of mental arithmetic task, the difficulty of mental arithmetic task can be controlled by time to induce different degrees of psychological stress. Adopting SBS algorithm and constructing BP neural network model and SVM classifier, the average recognition rate of the two classifiers combined with SBS and the two classifiers are above 80%. And selected to the two types of classifier contribution to the highest combination of characteristics, indicating that cardiac autonomic nervous activities can to a certain extent distinguish psychological stress, with the increase of stress. SDNN in the HRV index was significantly decreased and aHF was significantly increased. It can be inferred that cardiac sympathetic vagal nerve activity changes with the increase of the stress level. The sympathetic function is inhibited to some extent while the vagal function is relatively excited.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R318;TN911.7

【参考文献】