基于可穿戴设备的日常压力状态评估研究

发布时间：2018-05-16 10:22

本文选题：压力识别 + 可穿戴设备　；参考：《电子与信息学报》2017年11期

【摘要】：现代生活普遍压力较大,容易引起消极痛苦的应激,导致不良情绪甚至滋生各类慢性病。心理专家需要了解个体的压力状态,从而开展对应性心理疏导和治疗。传统心理学自评法存在一定的主观性;基于生理多导仪的压力状态评估法,受设备体积所限无法用于日常压力状态评估。针对上述问题,该文采用可穿戴式传感设备实时采集个体生理信号,利用心理和生理的伴生关系,对个体的心理压力进行长期实时评估。同时通过蒙特利尔影像应激实验(MIST)诱发出被试平静、轻微及高度压力3种压力状态,此实验范式同时包含认知负荷精神压力因素与社会评价心理压力因素,与日常真实生活更为接近。该文共采集39名健康被试的实验数据,通过对数据的特征值提取等预处理,结合随机森林算法对最优特征子集进行选择,采用支持向量机(SVM)分类算法对3种压力状态进行分类预测。实验结果表明,通过随机森林特征选择优化后的SVM分类,与通用的单一SVM分类算法相比,具有更好的分类识别效果,对3种压力状态的分类准确率可从78%提高至84%。
[Abstract]:The stress of modern life is high, which can lead to negative and painful stress, leading to bad mood and even various chronic diseases. Psychological experts need to understand the individual stress state, so as to develop corresponding psychological counseling and treatment. The traditional psychological self-evaluation method has some subjectivity, and the pressure state evaluation method based on physiological multi-conductors can not be used to evaluate the daily stress state due to the limitation of the equipment volume. To solve the above problems, the wearable sensing equipment is used to collect individual physiological signals in real time, and the long-term real-time evaluation of individual psychological pressure is carried out by using the relationship between psychology and physiology. At the same time, three kinds of stress states were induced by Montreal image stress test (MIST), which included cognitive load, mental stress factors and social evaluation psychological stress factors. Closer to everyday real life. In this paper, the experimental data of 39 healthy subjects were collected, and the optimal feature subset was selected by preprocessing the data, such as extracting the eigenvalue of the data, and combining with the stochastic forest algorithm. Support vector machine (SVM) classification algorithm is used to classify and predict three kinds of pressure states. The experimental results show that, compared with the general single SVM classification algorithm, the optimized SVM classification based on stochastic forest features has better classification and recognition effect, and the classification accuracy of the three pressure states can be improved from 78% to 84%.
【作者单位】：中国科学院电子学研究所;中国科学院大学;中国科学院心理研究所;
【基金】：国家自然科学基金(61302033) 北京市自然科学基金(Z160003) 国家重点研发计划(2016YFC1304302,2016YFC026502,2016YFC1303900)~~
【分类号】：R318.6;TP18

【相似文献】