Validation of Driver’s Cognitive Load on Driving Performance Using Spectral Estimation Based on EEG Frequency Spectrum
Validation of Driver’s Cognitive Load on Driving Performance Using Spectral Estimation Based on EEG Frequency Spectrum
复制标题
使用基于脑电图频谱的频谱估计来验证驾驶员的认知负荷对驾驶性能的影响
DOI：
10.1007/978-3-030-28505-0_5
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
S. Yaacob
中科院分区：
文献类型：
--
作者：
Mohamed Fredj;P. Krishnan;S. Yaacob
在过去的几年里，司机的嗜睡成为导致道路交通事故不断增加的一个突出因素，对道路安全造成了困扰。本研究提出了一种基于脑电图和功率谱的困倦和警觉识别方法，以评估静态驾驶模拟器中驾驶员的警觉水平。EEG数据库采用Karolinska嗜睡量表（KSS）和反应时间（RT）进行验证。使用监督学习分类器（MLNN、QSVM和KNN）评估频域功率谱密度（PSD）特征提取技术（周期图、Lomb-Scargle、Thompson多锥和Welch）。使用Lomb-Scargle PSD的MLNN准确率最高，为96.3%；使用周期图和Welch PSD特征集的QSVM和KNN准确率最低，均为62.2%。
Driver’s drowsiness becomes a prominent factor that causes the growing number of a road accident in the past few years and turns out to be perturbing for road safety. This research presents approaches for drowsiness and alertness recognition based on the electroencephalography (EEG) and power spectrum to evaluate the driver’s vigilance level in a static driving simulator. The EEG databases are validated using the Karolinska sleepiness scale (KSS) and reaction time (RT). Frequency-domain power spectral density (PSD) feature extraction techniques were evaluated (periodogram, Lomb-Scargle, Thompson multitaper, and Welch) with supervised learning classifiers (MLNN, QSVM, and KNN). The highest accuracy is attained from MLNN using Lomb-Scargle PSD with 96.3% and the minimum accuracy is attained from QSVM and KNN with both 62.2% using periodogram and Welch PSD features set respectively.