摘要
为了有效地评测人的驾驶精神疲劳状态,本文提出了一种基于核学习算法的精神疲劳分级方法。该方法首先用多变量自回归模型(MVAR)提取于前额、顶叶、枕叶共6个通道的多维脑电信号特征组成特征向量。然后用核主分量分析(KPCA)和优化支持向量机(SVM)对基于脑电信号(EEG)的驾驶精神疲劳进行分级。经过对3位受试者在3个状态下的驾驶精神疲劳进行分类,平均分类精度达到89.47%。分析显示,应用KPCA并结合优化SVM方法有效地降低了特征空间的维数,可实现较高精度的驾驶精神疲劳分级。
To effectively identify driving mental fatigue states, a new method based on kernel learning algorithm is presented. Firstly, multivariate autoregressive (MVAR) model is used to extract the feature vectors of six-channel electroencephalogram(EEG) signals from frontal, central and occipital electrods. Then, kernel principal component analysis (KPCA) and support vector machines (SVMs) with optimal parameters are proposed to classify driving mental fatigue. The method is used to classify three-level driving mental fatigue over 3 subjects. The average classification accuracy reaches to 89.47%. The result indicates that KPCA combined with optimal SVM can significantly reduce the dimensions of the feature vectors and obtain higher accuracy for classifying the driving mental fatigue.
出处
《数据采集与处理》
CSCD
北大核心
2009年第3期335-339,共5页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(30670534)资助项目
关键词
核主分量分析
支持向量机
多变量自回归模型
驾驶精神疲劳
脑电
kernel principal component analysis(KPCA)
support vector maehines(SVMs
multivariate autoregressive model
driving mental fatigue
electroencephalogram (EEG)