驾驶疲劳脑电信号的非线性动力学分析

【作者】 刘苗苗；

【导师】 艾玲梅；

【作者基本信息】 陕西师范大学 ， 计算机应用技术， 2011， 硕士

【摘要】 目前,驾驶疲劳已经成为引发交通事故的主要因素之一,给社会带来了巨大的经济损失,同时也造成了重大的人员伤亡,因此有效的驾驶疲劳检测方法,对预防疲劳驾驶、干预疲劳驾驶等具有非常重要的意义。已有的研究发现脑电信号能够准确反映人的生理状态,更被公认为是评价疲劳的“金标准”。大量学者将脑电信号用于驾驶疲劳的研究并取得了一定的成效,然而,这些研究大多都是从脑电信号的时域、频域等方面进行分析的,对于具有非线性特征的脑电信号,这些分析方法难以得到理想的效果。因此,本论文从非线性动力学角度来研究驾驶疲劳脑电信号,通过分析模拟驾驶过程中所采集的12名受试者的脑电信号,试图探索驾驶疲劳时的脑电信号特性。本论文的主要工作如下：首先,采用小波变换的阈值方法对欲分析的脑电信号进行降噪处理,以便提高分析的精确性。其次,先用相空间重构理论对疲劳和非疲劳两种驾驶状态的脑电信号进行相空间重构,然后计算12名受试者疲劳和非疲劳两种驾驶状态脑电信号的相关维数值和Lyapunov指数值,并对其进行分析,结果表明相关维数和Lyapunov指数在不同驾驶状态下的值均有明显的差异,都可用于区分不同的驾驶状态。最后,使用C0复杂度、近似熵、样本熵和多尺度熵方法计算12名受试者疲劳和非疲劳两种驾驶状态脑电信号的对应值,并对其进行分析,结果表明C0复杂度、近似熵、样本熵和多尺度熵在不同驾驶状态下的值都存在明显的差异,且近似熵比C0复杂度差异更明显,多尺度熵比近似熵和样本熵差异更明显。综上所述,脑电信号的相关维、Lyapunov指数、C0复杂度、近似熵、样本熵和多尺度熵都可以作为检测驾驶疲劳的指标,只是其检测结果存在着一定的差异。更多还原

【Abstract】 Currently, driving fatigue has been one of the main reasons of traffic accidents, bringing terrible economic lost and large scale of damages. In this case, detecting driving fatigue efficiently is of great importance to prevent and control driving fatigue.Researches show that electroencephalogram (EEG) could correctly reflect physical conditions and regarded as a gold standard for evaluating fatigue. A large number of scholars study the driving fatigue by EEG and made some improvements. However, these studies mainly focus on time domain, or frequency domain, which is not quite effective to non-linear signals like EEG. Therefore, this article take the angle of non-linear dynamic to analyze the EEG that colleted from 12 subjects in the process of a driving simulation, trying to study the characters of EEG during driving. The main work of this paper is:First, threshold of wavelet transform is used to denoise the EEG in order to raise the accuracy of the analysis.Second, phase space reconstruct theory is used to reconstruct the statement of fatigue and non-fatigue. Then calculate the different correlation dimension and Lyapunov index values of 12 subjects in vary driving statements. The analysis shows that correlation dimension and Lyapunov index value are both on different in different statements. They are available to distinguish different driving statements.At last, calculate the corresponding values of the subjects in different driving statements with c0 complexity, approximate entropy, sample entropy and multi-scale entropy. The calculations show that the results of all the methods used are different in different statements. Approximate entropy is more obvious than c0 complexity and the multi-scale entropy is more obvious than the approximate entropy and the sample entropy.Therefore, the correlation dimension, Lyapunov index, c0 complexity, approximate entropy, sample entropy and multi-scale entropy are available to be an indication of detecting driving fatigue, only different in the detecting results.更多还原