考虑初始情绪的个性化驾驶负荷状态评价

Initial Emotion-based Evaluation of the Personalized Driving Load State

利用驾驶人生理数据对驾驶人的负荷状态进行评价已成为交通心理学的研究热点,该方法通常需要采集驾驶人在静息状态的生理信号特征作为其负荷基准,因此负荷基准的提取将影响驾驶人状态评价结果的准确性。基于此,研究搭建驾驶模拟试验平台,招募15名志愿者开展驾驶模拟试验,设计不同任务诱导其产生3种程度的精神负荷,采集志愿者在不同负荷状态下的生理信号。基于试验数据研究驾驶人初始情绪对心电信号负荷基准值的影响,设计初始情绪提取处理方案,并提出基于个性化敏感生理特征的驾驶负荷评价方法。首先对原始心电信号进行滤波,修订心跳间隔异常值;其次计算心率变异性序列(HRV)时频域特征,利用线性回归对静息状态驾驶人情绪特征进行提取,并利用归一化处理消除初始情绪对驾驶人静息状态负荷的影响,完成驾驶人心电信号时域和频域特征的计算提取;最后采用Fisher-score特征选择算法完成不同驾驶人个体敏感特征的选取,并对个性化特征提取前后的驾驶人负荷评价结果进行对比分析。研究结果表明:研究设计的考虑初始情绪的个性化特征分类器可有效消除初始情绪对驾驶人静息状态负荷的影响,且提高了驾驶人负荷状态识别的准确率,可为进一步研究驾驶任务对驾驶人负荷状态的影响并改进车内外分心源的设计提供依据。

Using driver physiological data to evaluate drivers’ load status has become a major research focus in traffic psychology. This method requires the extraction of the drivers ’ physiological signal characteristics during the resting state as a load baseline,which affects the evaluation accuracy of the driver’s condition. In this study,a driving simulation experiment platform was established,and 15 volunteers were recruited to participate in the driving simulation experiments. Different tasks were designed to induce three levels of mental loads,and the physiological signals of the volunteers were collected under the different load conditions. Based on the experimental data,the influence of the drivers’ initial emotions on the load baseline value of the ECG signal was examined. Moreover,a scheme to process the extracted initial emotions was formulated,and a method to evaluate the driver’s load was established based on the personalized sensitive physiological characteristics. First,the original ECG signal was filtered,and the abnormal value of the heartbeat interval was revised. Second,the timefrequency characteristics of the heart rate variability( HRV) were determined,and the emotional characteristics of the drivers in the resting state were extracted by linear regression. Subsequently,the influence of the initial emotion on the resting state load of the driver was eliminated by normalization processing,and the time-domain and frequency-domain characteristics of the driver’s ECG signal were calculated and extracted. Finally,the Fisher-score feature selection algorithm was used to select the sensitive features of the different drivers. The comparison and analysis of the driver load evaluation results before and after the personalized feature extraction indicated that the proposed personalized feature classifier could effectively eliminate the influence of the initial emotion on the driver’s resting state load.Moreover,this approach could enhance the accuracy of the driver’s load state recognition. The findings can provide a basis for further research on the influence of driving tasks on the drivers’ load states and help enhance the design of internal and external distractors.