吉林大学学报(工学版) ›› 2020, Vol. 50 ›› Issue (1): 121-131.doi: 10.13229/j.cnki.jdxbgxb20180918

• 交通运输工程·土木工程 • 上一篇    

基于ROC曲线和驾驶行为特征的驾驶愤怒强度判别阈值

万平1,2(),吴超仲2,马晓凤2()   

  1. 1. 华东交通大学 交通运输与物流学院,南昌 330013
    2. 武汉理工大学 智能交通系统研究中心,武汉430063
  • 收稿日期:2018-09-04 出版日期:2020-01-01 发布日期:2020-02-06
  • 通讯作者: 马晓凤 E-mail:pingw04@163.com;maxiaofeng@whut.edu.cn
  • 作者简介:万平(1984-),男,博士研究生. 研究方向:驾驶行为与安全辅助驾驶. E-mail:pingw04@163.com
  • 基金资助:
    国家自然科学基金项目(51775396);江西省社会科学规划青年博士基金项目(17BJ42);江西省教育厅科学技术研究项目(GJJ180359)

Discriminating threshold of driving anger intensity based on driving behavior features by ROC curve analysis

Ping WAN1,2(),Chao-zhong WU2,Xiao-feng MA2()   

  1. 1. School of Transportation and Logistics, East China Jiaotong University, Nanchang 330013, China
    2. Intelligent Transport Systems Research Center, Wuhan University of Technology, Wuhan 430063, China
  • Received:2018-09-04 Online:2020-01-01 Published:2020-02-06
  • Contact: Xiao-feng MA E-mail:pingw04@163.com;maxiaofeng@whut.edu.cn

摘要:

为了对“路怒症”进行分级预警,提出了一种基于驾驶行为特征的愤怒强度判别阈值求解方法。通过在交通繁忙路段开展限时实车试验获得不同愤怒强度下的驾驶行为数据。基于方差分析选用方向盘转动角速度(SWRAR)、加速踏板踩踏速度(SGP)、横向加速度(LA)作为驾驶愤怒判别指标。运用接受者操作特征(ROC)曲线分析方法求得4种愤怒强度的最佳阈值范围分别为:正常状态0≤SWRAR<1.886 3,0≤SGP<0.397 4, 0≤LA<0.136 2;低强度愤怒1.886 3≤SWRAR<2.324 1,0.397 4≤SGP<0.505 4,0.136 2≤LA<0.155 9;中强度愤怒2.324 1≤SWRAR<3.083 5,0.505 4≤SGP<0.614 2,0.155 9≤LA<0.175 4;高强度愤怒SWRAR≥3.083 5,SGP≥0.614 2,LA≥0.175 4。验证结果表明,此驾驶行为特征阈值的准确率可达79.42%。本文研究结果可为驾驶愤怒分级预警系统的开发提供理论支撑。

关键词: 道路工程, 驾驶愤怒阈值, ROC曲线, 驾驶行为, 交通安全

Abstract:

In order to make effective graded warning for road rage, an optimal threshold determination method for discriminating driving anger intensity based on driving behavior features is proposed in this study. Different driving behavior data were acquired in driving states with different anger intensity by conducting timed experiments in busy traffic sections. The steering wheel rotation angel rate (SWRAR), speed of gas pedaling (SGP), and lateral acceleration (LA) were selected as three effective driving behavior features for discriminating different anger intensity by analysis of variance. A method of receiver operating characteristic (ROC) curve analysis was introduced to determine the optimal discrimination thresholds of SWRAR, SGP and LA. For normal state the ranges of the thresholds are 0≤SWRAR<1.8863, 0≤SGP<0.397 4, 0≤LA<0.136 2; for low anger state the ranges are 1.886 3≤SWRAR<2.324 1, 0.397 4≤SGP<0.505 4, 0.136 2≤LA<0.155 9; for moderate anger state the ranges are 2.324 1≤SWRAR<3.083 5, 0.505 4≤SGP<0.614 2, 0.155 9≤LA<0.175 4; and for high anger state the ranges are SWRAR 3.083 5, SGP 0.614 2, LA 0.175 4. The verification performances show that the accuracy of the optimal thresholds for discriminating different anger intensity can reach 79.42%. The results can provide theoretical foundation for developing multilevel warning systems for driving anger based on the optimal discrimination thresholds.

Key words: road engineering, driving anger discrimination, receiver operating characteristic (ROC)curve, driving behavior, traffic safety

中图分类号: 

  • U491

图1

实车试验路线"

图2

试验中的道路事件"

图3

实车试验用车"

图4

高清监测摄像头"

图5

被试自我报告与观察者评价相关系数"

图6

驾驶愤怒样本强度分布"

表1

4种愤怒强度下的驾驶行为特征的描述性统计"

驾驶行为指标 驾驶愤怒强度
SGP/(m?s-1) 0.289(0.124) 0.458(0.171) 0.579(0.211) 0.656(0.247)
SBP/(m?s-1) 0.218(0.062) 0.224(0.066) 0.228(0.068) 0.233(0.071)
SWRAR/[(°)?s-1] 1.316(0.142) 2.083(0.167) 2.658(0.191) 3.404(0.232)
FA/(m?s-2) 0.469(0.121) 0.611(0.169) 0.736(0.197) 0.863(0.221)
LA/(m?s-2) 0.127(0.036) 0.148(0.042) 0.164(0.051) 0.188(0.057)

表2

4种愤怒强度下的驾驶行为特征方差检验结果"

驾驶行为指标 平方和 df 均方 F 显著性
SGP 组间 32.624 31 3 10.874 77 372.158 39 0.034
组内 51.720 83 1770 0.029 22
总数 84.345 14 1773
SBP 组间 0.046 29 3 0.015 43 3.624 64 0.126
组内 7.534 47 1770 0.004 26
总数 7.580 76 1773
SWRAR 组间 865.531 25 3 288.510 42 10042.519 16 0.028
组内 50.850 13 1770 0.028 73
总数 916.381 38 1773
FA 组间 32.047 02 3 10.682 34 404.607 89 0.083
组内 46.731 02 1770 0.026 40
总数 78.778 04 1773
LA 组间 0.707 95 3 0.235 98 126.299 67 0.042
组内 3.307 14 1770 0.001 87
总数 4.015 09 1773

图7

ROC曲线坐标系统及其评价指标示图"

图8

愤怒等级1、3、5的SWRAR判别检测的ROC曲线"

图9

愤怒等级1、3、5的SGP判别检测的ROC曲线"

图10

愤怒等级1、3、5的LA判别检测的ROC曲线"

表3

SGP、SWRAR与LA判别不同愤怒等级的AUC统计分析"

愤怒等级 判别指标 AUC Std.Error Asymptotic Sig. 95%置信区间
上限 下限
SWRAR 0.821 9 0.031 4 0.034 0.769 7 0.842 8
等级1 SGP 0.835 4 0.035 1 0.029 0.752 4 0.853 1
LA 0.792 5 0.028 7 0.042 0.704 8 0.821 6
SWRAR 0.831 4 0.039 2 0.039 0.794 3 0.862 5
等级3 SGP 0.852 7 0.042 6 0.023 0.816 7 0.881 3
LA 0.822 7 0.032 4 0.023 0.790 4 0.853 9
SWRAR 0.853 6 0.041 2 0.037 0.810 3 0.884 9
等级5 SGP 0.873 4 0.052 8 0.018 0.821 8 0.904 6
LA 0.824 5 0.037 3 0.046 0.785 4 0.862 7

表4

SGP、SWRAR与LA判别不同愤怒等级的最佳阈值以及判别准确率"

愤怒等级 判别指标 TPR FPR 最佳判别阈值
等级1 SWRAR 0.8276 0.2142 1.8863
SGP 0.8043 0.2015 0.3974
LA 0.7625 0.1438 0.1362
等级3 SWRAR 0.7489 0.1632 2.3241
SGP 0.8673 0.2147 0.5054
LA 0.8137 0.1847 0.1559
等级5 SWRAR 0.7827 0.1524 3.0835
SGP 0.8344 0.1781 0.6142
LA 0.7538 0.1395 0.1754

表5

4种驾驶愤怒强度的最佳判别阈值范围"

驾驶行为指标 正常 低强度愤怒 中强度愤怒 高强度愤怒
anger level<1 1≤anger level<3 3≤anger level<5 anger level≥5
SWRAR [0,1.886 3) [1.886 3,2.324 1) [2.324 1,3.083 5) [3.083 5,+∞)
SGP [0,0.397 4) [0.397 4,0.505 4) [0.505 4,0.614 2) [0.614 2,+∞)
LA [0,0.136 2) [0.136 2,0.155 9) [0.155 9,0.175 4) [0.175 4,+∞)

图11

基于驾驶行为特征阈值的驾驶愤怒强度判别流程"

表6

基于驾驶行为特征阈值融合的驾驶愤怒强度判别模型的测试结果 %"

愤怒强度 TPR PPA F1 Ac
正常驾驶 84.27 89.65 86.87 79.42
低强度愤怒 75.18 75.83 75.51
中等强度愤怒 78.24 69.39 73.55
高强度愤怒 82.35 73.46 77.65
1 吴超仲,雷虎 . 汽车驾驶愤怒情绪研究现状与展望[J]. 中国安全科学学报, 2010, 20(7): 3-8.
Wu Chao-zhong , Lei Hu . Review on the study of motorists’ driving anger[J]. China Safety Science Journal, 2010, 20(7):3-8.
2 Kotinas I . Traffic safety facts: a compilation of motor vehicle crash data from the fatality analysis reporting system and the general estimates system[R]. The U.S. National Highway Traffic Safety Administration, 2011.
3 雷虎 . 愤怒情绪下的汽车驾驶行为特征及其对交通安全的影响研究[D]. 武汉: 武汉理工大学智能交通系统研究中心, 2011.
Lei Hu . The characteristics of angry drivingbehaviors and its effects on traffic safety [D]. Wuhan: Intelligent Transport Systems Research Center, Wuhan University of Technology, 2011.
4 Feng Zhong-xiang , Lei Ye-wei , Liu Hong-chao , et al . Driving anger in China: a case study on professional drivers[J]. Transportation Research Part F: Traffic Psychology and Behavior, 2016, 42: 255-266.
5 Dahlen E R , Martin R C , Ragan K , et al . Driving anger, sensation seeking, impulsiveness, and boredom proneness in the prediction of unsafe driving[J]. Accident Analysis and Prevention, 2005, 37(2): 341-348.
6 张迪, 万柏坤, 明东 . 基于生理信号的情绪识别研究进展[J]. 生物医学工程学杂志, 2015, 32(1): 229-234.
Zhang Di , Wan Bai-kun , Ming Dong . Research progress on emotion recognition based on physiological signals[J]. Journal of Biomedical Engineering, 2015, 32(1): 229-234.
7 Paschero M , Vescovo G , Benucci L , et al . Real time classifier for emotion and stress recognition in a vehicle driver[C]∥IEEE International Symposium on Industrial Electronics,Hangzhou,China, 2012: 1690-1695.
8 Kamaruddin N , Wahab A . Driver behavior analysis through speech emotion understanding[C]∥IEEE Intelligent Vehicles Symposium University of California, San Diego, USA, 2010: 238-243.
9 Katsis C D , Goletsis Y , Rigas G , et al . A wearable system for the affective monitoring of car racing drivers during simulated conditions[J]. Transportation Research Part C: Emerging Technologies, 2011, 19(3): 541-551.
10 Rebolledo-Mendez G , Reyes A , Paszkowicz S , et al . Developing a body sensor network to detect emotions during driving[J]. IEEE Transactions on Intelligent Transportation Systems, 2014, 15(4): 1850-1854.
11 Kessous L , Castellano G , Caridakis G . Multimodal emotion recognition in speech-based interaction using facial expression, body gesture and acoustic analysis[J]. Journal on Multimodal User Interfaces, 2010, 3(1): 33-48.
12 Cai H , Lin Y , Mourant R . Study on driver emotion in driver-vehicle-environment systems using multiple networked driving simulators[C]∥Driving Simulation Conference, Iowa City, Iowa, America, 2007: 1-8.
13 万平,吴超仲,林英姿,等 . 基于置信规则库的驾驶人愤怒情绪识别模型[J]. 交通运输系统工程与信息, 2015, 15(5): 96-102.
Wan Ping , Wu Chao-zhong , Lin Ying-zi , et al . A recognition model of driving anger based on belief rule base[J]. Journal of Transportation Systems Engineering and Information Technology, 2015, 15(5): 96-102.
14 刘鹏 . 融合面部表情和语音的驾驶员路怒症识别方法研究[D]. 镇江:江苏大学计算机科学与通信工程学院,2017.
Liu Peng . Driver road rage recognition by combining facial expression and speech[D]. Zhenjiang: School of Computer Science and Telecommunication Engineering, Jiangsu University, 2017.
15 Herrero-Fernánde D . Psychometric adaptation of the driving anger expression inventory in a Spanish sample: differences by age and gender[J]. Transportation Research Part F: Psychology and Behavior, 2011, 14(4): 324-329.
16 万平,吴超仲,林英姿,等 . 基于驾驶行为多元时间序列特征的愤怒驾驶状态检测[J]. 吉林大学学报: 工学版, 2017, 47(5): 1426-1435.
Wan Ping , Wu Chao-zhong , Lin Ying-zi , et al . Driving anger detection based on multivariate time series features of driving behaviors[J]. Journal of Jilin University(Engineering and Technology Edition), 2017, 47(5): 1426-1435.
17 Chen Zhi-jun , Wu Chao-zhong , Zhong Ming , et al . Identification of common features of vehicle motion under drowsy/distracted driving a case study in Wuhan China[J]. Accident Analysis and Prevention, 2015, 81: 251-259.
18 陈卫中,倪宗瓒,潘晓平,等 .用ROC曲线确定最佳临界点和可疑值范围[J]. 现代预防医学, 2005, 32(7) : 729-731.
Chen Wei-zhong , Ni Zong-zan , Pan Xiao-ping , et al . Receiver operating characteristic curves to determine the optimal operating point and doubtable value interval[J]. Modern Preventive Medicine, 2005, 32(7): 729-731.
19 Bechtel T , Capineri L , Windsor C , et al . Comparison of ROC curves for landmine detection by holographic radar with roc data from other methods[C]∥IEEE 8th International Workshop on Advanced Ground Penetrating Radar,Florence,Italy,2015:1-4.
[1] 王芳,李晓光,郭慧,胡佳. 基于驾驶员视觉兴趣区的沙漠草原公路曲线间直线段线形指标优化[J]. 吉林大学学报(工学版), 2020, 50(1): 114-120.
[2] 狄胜同,贾超,乔卫国,李康,童凯. 橡胶集料混凝土细观损伤特性的加载速率效应[J]. 吉林大学学报(工学版), 2019, 49(6): 1900-1910.
[3] 张云龙,周刘光,王静,吴春利,吕翔. 冻融对粉砂土力学特性及路堤边坡稳定性的影响[J]. 吉林大学学报(工学版), 2019, 49(5): 1531-1538.
[4] 彭勇,高华,万蕾,刘贵应. 沥青混合料劈裂强度影响因素数值模拟[J]. 吉林大学学报(工学版), 2019, 49(5): 1521-1530.
[5] 江亮,贺宜. 电动两轮车风险驾驶行为及事故影响因素分析[J]. 吉林大学学报(工学版), 2019, 49(4): 1107-1113.
[6] 李晓珍,柳俊哲,戴燕华,贺智敏,巴明芳,李玉顺. 碳化作用下水泥浆内亚硝酸根离子的含量分布[J]. 吉林大学学报(工学版), 2019, 49(4): 1162-1168.
[7] 于天来,李海生,黄巍,王思佳. 预应力钢丝绳加固钢筋混凝土梁桥抗剪性能[J]. 吉林大学学报(工学版), 2019, 49(4): 1134-1143.
[8] 黄晓明,曹青青,刘修宇,陈嘉颖,周兴林. 基于路表分形摩擦理论的整车雨天制动性能模拟[J]. 吉林大学学报(工学版), 2019, 49(3): 757-765.
[9] 王静,吕翔,曲肖龙,钟春玲,张云龙. 路基土抗剪强度与化学及矿物成分的关系[J]. 吉林大学学报(工学版), 2019, 49(3): 766-772.
[10] 李伊,刘黎萍,孙立军. 沥青面层不同深度车辙等效温度预估模型[J]. 吉林大学学报(工学版), 2018, 48(6): 1703-1711.
[11] 念腾飞, 李萍, 林梅. 冻融循环下沥青特征官能团含量与流变参数灰熵分析及微观形貌[J]. 吉林大学学报(工学版), 2018, 48(4): 1045-1054.
[12] 臧国帅, 孙立军. 基于惰性弯沉点的刚性下卧层深度设置方法[J]. 吉林大学学报(工学版), 2018, 48(4): 1037-1044.
[13] 宫亚峰, 申杨凡, 谭国金, 韩春鹏, 何钰龙. 不同孔隙率下纤维土无侧限抗压强度[J]. 吉林大学学报(工学版), 2018, 48(3): 712-719.
[14] 程永春, 毕海鹏, 马桂荣, 宫亚峰, 田振宏, 吕泽华, 徐志枢. 纳米TiO2/CaCO3-玄武岩纤维复合改性沥青的路用性能[J]. 吉林大学学报(工学版), 2018, 48(2): 460-465.
[15] 马晔, 尼颖升, 徐栋, 刁波. 基于空间网格模型分析的体外预应力加固[J]. 吉林大学学报(工学版), 2018, 48(1): 137-147.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!