基于SVR模型的驾驶简单反应时间预测方法

发布时间：2018-05-18 07:34

本文选题：交通工程 + 简单反应时间　；参考：《中国公路学报》2017年04期

【摘要】：为准确预测驾驶人对突发事件的简单反应时间,并为构建自适应式危险驾驶状态预警系统奠定基础,提出了一种基于实时脑电图(EEG)信号的驾驶人简单反应时间预测方法。首先,通过快速傅里叶变换(FFT)对EEG信号进行特征参数提取,作为驾驶人简单反应时间的客观预测指标。在此基础上,基于支持向量回归(SVR)建立驾驶人对突发事件简单反应时间的预测模型。最后采用20名驾驶人连续驾驶4h的EEG信号与反应时间数据,对该模型予以试算。研究结果表明:3项脑电特征参数(θ,α,β)与反应时间均具有显著相关性,其中脑电特征参数α的相关性最显著,为SVR模型对简单反应时间进行预测提供了客观预测指标;分别采用径向基函数(RBF)、Polynomial函数、Sigmoid函数作为核函数构建SVR模型对简单反应时间进行预测时,所得预测结果中采用RBF函数所产生的各项误差均低于其他2项函数,表明采用RBF函数为核函数的SVR模型预测精度最优,其预测准确率达到80%以上。
[Abstract]:In order to accurately predict the simple reaction time of the driver to the unexpected events and lay the foundation for the construction of adaptive dangerous driving state early warning system, a simple reaction time prediction method based on real time electroencephalograph (EEG) signal is proposed. First, the characteristic parameters of EEG signal are extracted by fast Fu Li Ye Bianhuan (FFT). On this basis, based on the support vector regression (SVR), a driver's prediction model for the simple response time of a driver is built on this basis. Finally, the model is tested with the EEG signal and the reaction time data of 20 drivers driving the 4H continuously. The results show that 3 EEG characteristic parameters are used. The number (theta, alpha, beta) has a significant correlation with the reaction time, and the correlation of the EEG parameter alpha is the most significant. The SVR model provides an objective predictor for the prediction of the simple reaction time. The RBF, Polynomial function and Sigmoid function are used as the kernel function to build the SVR model for the simple reaction time. The error generated by the RBF function in the prediction results is lower than the other 2 functions. It shows that the prediction accuracy of the SVR model with RBF function as a kernel function is the best, and the accuracy of the prediction is above 80%.
【作者单位】：西南交通大学交通运输与物流学院;西南交通大学综合交通运输智能化国家地方联合工程实验室;
【基金】：国家自然科学基金项目(51108390) 国家重点研发计划项目(2016YFC0802209)
【分类号】：U491.254

【参考文献】