基于贝叶斯网络的驾驶疲劳程度识别模型

国内外学者大多采用单一类型指标对驾驶疲劳程度进行判断。为克服单一指标检测的不稳定性，构建基于贝叶斯网络的驾驶疲劳程度识别模型。将驾驶环境属性、驾驶人个体属性和原始疲劳属性作为模型输入层变量。选择脑电指标、心电指标、眼动指标、驾驶绩效指标作为模型输出层变量。将清醒、轻度疲劳、重度疲劳三种驾驶疲劳程度作为隐含层变量。采用模拟驾驶方法进行实验，得到不同实验对象各个时刻不同疲劳程度的概率。将利用单一指标和贝叶斯网络模型得到的驾驶人疲劳程度与主观疲劳测评结果进行对照，证明贝叶斯网络模型不仅能消除单一指标失效时产生的误判和漏判，而且可提高识别的准确性。

Driving Fatigue Recognition Model Based on Bayesian Network

Scholars both at home and abroad commonly investigate driving fatigue using single type indicators. To overcome the instability of single indicator detection, this paper develops a driving fatigue recognition model based on Bayesian Network. It takes environmental attribute, individual attribute of drivers and original fatigue attribute as the variables of the input layer of the model;regards θ/β index, SDNN index, PERCLOS index and SDS index as the variables of the output layer;and uses soberness, mild fatigue, and severe fatigue as the variables of the hidden layer. Based on the experiment with driving simulation meth-od, the probability of different degrees of fatigue by different experimental subjects at different times is obtained. The paper then compares driving fatigue gained through single index and Bayesian Network model with the subjective fatigue evaluation results, showing that the Bayesian Network model could not only eliminate misjudgment caused by invalidity of single index, but also increase the accuracy of recognition.