| 引用本文: | 王琳虹,李世武,高振海,冀秉魁.基于粒子群优化与支持向量机的驾驶员疲劳等级判别[J].哈尔滨工业大学学报,2014,46(12):102.DOI:10.11918/j.issn.0367-6234.2014.12.017 |
| WANG Linhong,LI Shiwu,GAO Zhenhai,JI Bingkui.A driver fatigue level recognition model based on particle swarm optimization and support vector machine[J].Journal of Harbin Institute of Technology,2014,46(12):102.DOI:10.11918/j.issn.0367-6234.2014.12.017 |
|
| 摘要: |
| 为客观、准确地判别驾驶员的疲劳程度,采用多项驾驶员生理指标、基于粒子群与支持向量机(SVM)算法建立驾驶疲劳等级判别模型,首先将驾驶员疲劳状态分为清醒、轻度疲劳、重度疲劳和睡意4个等级,然后将驾驶员的心电RR间期标准差、心率均值、呼吸潮气量、脑电的α波、β波和δ波功率谱密度积分等作为SVM的输入变量,驾驶疲劳等级作为输出变量,引入粒子群算法优化SVM的惩罚系数和核函数参数对判别模型进行标定,采用吉珲高速公路上的实车实验数据对模型有效性进行验证. 结果表明:本模型对4项疲劳等级的判别准确率均高于85％,对于驾驶员疲劳预警具有重要意义．通过对模型各个输入变量的敏感性分析,证明基于多项生理指标的疲劳判别较基于单生理指标的疲劳判别更加有效． |
| 关键词: 驾驶员生理指标 疲劳判别 支持向量机 粒子群 敏感性分析 |
| DOI：10.11918/j.issn.0367-6234.2014.12.017 |
| 分类号:U491.2 |
| 基金项目:国家自然科学基金(51308251); 中国博士后科学基金(2013M541306)． |
|
| A driver fatigue level recognition model based on particle swarm optimization and support vector machine |
|
WANG Linhong1,2, LI Shiwu1, GAO Zhenhai2, JI Bingkui1
|
|
(1. College of Transportation, Jilin University, 130025 Changchun, China; 2. College of Automotive Engineering, Jilin University, 130025 Changchun, China)
|
| Abstract: |
| To recognize driver′s fatigue level accurately and objectively, a driver fatigue level recognition model that employs multiple psychological features was developed based on particle swarm optimization (PSO) and support vector machine (SVM). Firstly, the driver fatigue was divided into four levels, which were alert, mild fatigue, deep fatigue and drowsiness. Then alpha rhythm, beta rhythm, delta rhythm, mean of heart rate, and standard deviation of R-R interval were selected as input variables of the SVM model. The PSO was introduced into the model to optimize the penalty parameter and kernel function parameter of SVM. Experimental data collected in Ji-Hun freeway was used to validate the effectiveness of the recognition model. Results show that the recognition precision of the four fatigue levels are higher than 85％. Sensitive analysis of the mode is also conducted and the results prove that the model using multiple features outperforms the model using fewer features. |
| Key words: driver′s psychological features fatigue recognition support vector machine particle swarm optimization sensitive analysis |
