基于支持向量机的皮电信号心理应激识别

论文总字数：31865字

摘 要

心理学研究表明，适当的压力能够诱发积极向上的情绪、激发人们的潜力、提高工作效率，但是如果这种刺激超过了人体所能承受的一个界限，就会引起个体生理和心理上的一些疾病。当人体感受到这种心理应激时，其生理上会发生一系列的变化，但并非所有的生理指标都与心理应激密切相关。

皮肤电信号（GSR，galvanic skin response）是与心理应激密切相关的典型生理信号之一，它不易被掩盖且采集方便，对心理应激状态的改变反应灵敏且不受主观因素的影响，是心理应激状态分类识别的良好指标。在众多的分类器算法中， SVM算法在处理小样本高维数据时表现优异，分类识别率较高。本课题将利用MIT 公布的真实驾驶心理应激数据库，提取有效的皮肤电信号进行信号预处理、特征提取与归一化处理，应用Python编程语言对基于SVM算法的心理应激识别分类器进行建模，同时将采用随机森林算法与xgboost算法与SVM算法进行比较。

结果表明，通过网格搜索法，得到RBF内核的最优参数解为：C=8, gamma=0.0077。基于SVM算法（RBF内核）的心理应激等级分类器的识别率为79.01%，基于随机森林算法与xgboost算法的识别准确率相差很小均为72.97%。该结果说明，基于SVM算法的心理应激等级分类器的识别效果较为理想且SVM算法在与其他两种算法的比较中表现最好。本课题实现了基于单模态生理信号（皮肤电信号）的心理应激分类模型构建，其性能较为理想、模型简单、计算成本低，在未来有较大的实际应用价值。

关键词：心理应激，支持向量机，皮肤电信号

ABSTRACT

Psychological research shows that appropriate stress can induce positive emotions, stimulate people's potential, and promote people's learning and work efficiency, but if this stimulus exceeds a limit that the human body can bear, it will cause the individual's physical and psychological diseases. In medical medicine, psychological stress is also named psychological stress. When the human body feels this sort of psychological stress, it will suffer a series of changes in its physiology, but not all physiological indicators are closely related to psychological stress.

The galvanic skin response is one of the typical physiological signals closely related to psychological stress. It is not easy to be covered and easy to collect. It is sensitive to the change of psychological stress state and is not affected by subjective factors. It is an effective identification of psychological stress state. Many classical algorithms are suitable for the processing of physiological signals. The SVM algorithm performs well in processing small-sample and high-dimensional data, and has many kernel functions for us to flexibly select the appropriate kernel. In summary, this paper will be based on the driver's galvanic skin response collected by MIT experimenters in a real driving environment, then do the pre-processing and feature extraction of the physiological signals, at last use the Python programming language to model the SVM algorithm for psychological stress recognition. At the same time, Random Forest algorithm and xgboost algorithm are also used to compare with SVM algorithm.

The results show that the optimal parameter solution based on RBF kernel is found by grid search method: C=8, gamma=0.0077. The recognition rate of the psychological stress state classifier based on the SVM algorithm RBF kernel is 79.01%, and the difference between the recognition accuracy based on the random forest algorithm and the xgboost algorithm is 72.97%. In summary, the SVM algorithm performs best in the comparison of the other two algorithms. Secondly, in comparison with multimodal physiological signals, the recognition rate based on SVM algorithm for single-mode skin electrical signals is ideal and effective. The algorithm model is simple. The computational cost is low, and in the future, it will be of great significance for the development and practical application of equipment based on single mode.

KEYWORDS: psychological stress, support vector machine, galvanic skin response

目 录

第一章 绪论 1

1.1 研究背景与意义 1

1.2 国内外研究现状 2

1.3 本文研究内容 4

第二章 皮肤电信号获取与预处理 6

2.1 皮肤电信号简介 6

2.2 皮肤电信号获取 8

2.3 皮肤电信号去噪处理 11

2.4 皮肤电信号特征提取 12

2.5 本章小结 15

第三章 基于支持向量机算法的心理应激分类器设计 16

3.1 支持向量机原理与简介 16

3.2 SVM分类器算法实现 22

3.2.1 Python简介与其优势 22

3.2.2 Python编程实现 23

3.3 本章小结 28

第四章 心理应激分类的结果分析 29

4.1 分类器算法的程序结果 29

4.2 支持向量机与其他算法的分类效果比较 31

4.3 本章小结 32

第五章 工作总结与展望 33

5.1 工作总结 33

5.2 工作展望 33

参考文献 35

致 谢 37

第一章 绪论

剩余内容已隐藏，请支付后下载全文，论文总字数：31865字