基于工作记忆模型的脑力疲劳检测方法研究

发布时间：2018-01-19 13:24

本文关键词： 脑力疲劳脑电回归分析跨时间跨个体　出处：《天津大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着社会发展,人们生活节奏的加快、工作压力的增加,在如军事作业人员、驾驶员、外科手术医师、值夜班者等人群中,脑力疲劳的发生比例也随之增加。而严重的脑力疲劳会对人们的工作和生活产生极大的负面影响,甚至威胁人类的生命财产安全。因此,对脑力疲劳的检测是十分重要的,它能够为疲劳预警提供依据,帮助后续疲劳对抗措施的有序进行,从而减轻甚至消除脑力疲劳带来的危害。本论文设计了一种新型工作记忆任务(N-back)实验,将原本单一考察某一方面记忆的N-back划分为同时调用空间、颜色、形状记忆的新实验,并根据被试者训练情况调节负荷强度,从而更有效的诱发被试者脑力疲劳;在实验的同时,记录被试者脑电(electroencephalography,EEG)和N-back实验结果数据,并加入被试疲劳量表的填写;根据采集到的样本,对数据进行处理,筛选导联及特征,最终建立脑力疲劳预测回归模型;通过对脑电数据进行连续量化标注,比较了使用不同脑力疲劳程度标注建立的脑力疲劳模型优劣;此外,验证了脑力疲劳模型的跨时间、跨个体预测的正确性;并针对脑力疲劳标注不统一的现象,探讨脑力疲劳融合标注的效果。得到的主要结论有:以拟合优度检验为判据,使用搜索技术结合交叉验证筛选用于脑力疲劳预测模型的导联和特征,可建立脑力疲劳预测的回归模型;在对样本进行标注时,使用等效实验时长比等效正确率效果更好,前者最终建立的脑力疲劳预测回归模型其R2能达到0.9以上,脑电特征对脑力疲劳程度的解释性较强;对模型进行分析发现,建立的预测回归模型具有一定的跨时间、跨个体正确性,且不具有性别差异,但模型并不十分稳定,仍需要进行时间校正、个体校正,以及优化脑力疲劳的标注;对脑力疲劳融合标注的研究结果也表明,可使用等效正确率比值结合等效实验时长融合标注脑力疲劳程度。
[Abstract]:With the development of society, the pace of life of people, the increase of work pressure, such as military operators, drivers, surgeons, night shift and other people. The proportion of mental fatigue also increases. Serious mental fatigue will have a great negative impact on people's work and life, and even threaten the safety of human life and property. The detection of mental fatigue is very important. It can provide the basis for fatigue warning and help the follow-up fatigue countermeasures to proceed in an orderly manner. In order to reduce or even eliminate the harm caused by mental fatigue, this paper designed a new working memory task N-back-based experiment. N-back, which was originally used to examine a single aspect of memory, was divided into a new experiment of simultaneous transfer space, color and shape memory, and the load intensity was adjusted according to the training situation of the subjects. Thus more effectively induced mental fatigue; At the same time, the data of electroencephalography (EGG) and N-back test were recorded, and the participants were added to the fatigue scale. According to the collected samples, the data were processed, the lead and characteristics were screened, and the regression model of mental fatigue prediction was established. Through continuous quantification of EEG data, the advantages and disadvantages of mental fatigue model with different mental fatigue degree tagging were compared. In addition, the validity of cross-time and cross-individual prediction of mental fatigue model is verified. Aiming at the phenomenon of inconsistency of mental fatigue labeling, the effect of mental fatigue fusion labeling is discussed. The main conclusions are as follows: the goodness of fit test is taken as the criterion. The regression model of mental fatigue prediction can be established by using search technology and cross-validation to screen the leads and characteristics of the prediction model. In the labeling of samples, the length of the equivalent experiment is better than the equivalent correct rate. The R2 of the regression model of mental fatigue prediction established by the former can reach 0.9 or more. The characteristics of EEG explained the degree of mental fatigue more strongly. Analysis of the model found that the established predictive regression model has a certain cross-time, cross-individual correctness, and no gender difference, but the model is not very stable, still need to carry out time correction, individual correction. And optimizing the marking of mental fatigue; The results of the research on mental fatigue fusion labeling also show that the degree of mental fatigue can be marked by using the ratio of equivalent correct rate combined with the equivalent experiment.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R741.044

【参考文献】