摘要
为了有效判别驾驶员的疲劳驾驶状态,本文利用模拟驾驶器开展驾驶实验,采集了20名驾驶员在疲劳状态和正常状态的实验数据;为了提取能表征驾驶员疲劳和正常驾驶状态时的行为特征,本文对获取的速度、方向盘转角和车辆横向位置的样本熵进行了分析,最终提取了该三类参数的样本熵作为疲劳驾驶的有效特征组;构建了基于BP神经网络的驾驶员疲劳驾驶判别算法,并采用测试集样本对构建的算法进行验证.实验结果表明:该算法对于驾驶员疲劳驾驶检测的准确率较好、运行时间较短、具有较好的鲁棒性和实用性.
In order to effectively detect the fatigue driving behavior,a driving experiment was conducted in a driving simulator. The driving states were divided into 2 levels : fatigue state and normal state. A total of 20 drivers participated in the experiment. In order to extract the features that can effectively display the driver's driving behaviors,this paper compared the sample entropy of speed, steer and LP parameters,and found these features can measure the performance of the driving behavior well and can be used to build the classifier. Finally, a classifier based on BP was established to detect the fatigue driving,and a test set was used to verify this classifier. The results show that the classifier based on BP has a better detection accuracy,shorter running time and better robustness and practicability.
出处
《道路交通与安全》
2016年第6期21-24,共4页
Road Traffic & Safety
基金
交通运输部2014年度科技项目(批准号:2014364222110)
关键词
交通安全
驾驶行为
疲劳驾驶
BP神经网络
样本熵
traffic safety
driving behavior
fatigue driving
BP neural network
sample entropy