摘要
驾驶疲劳的产生是渐进的动态生成过程,基于隐马尔可夫模型(hidden Markov model,HMM)的相关研究需首先确定模型训练初值,且训练过程易陷入局部最优.基于此,通过在HMM训练过程中引入粒子群优化(particle swarm optimization,PSO)算法对训练过程存在的上述问题进行了改进,并结合驾驶疲劳状态典型数据集对所提出的改进方法和前向后向算法(forward-backward(BW)algorithm)进行了详细对比.实验及分析测试结果表明,所提出的改进方法在驾驶疲劳预测结果准确性和稳定性上都优于BW算法.
The generation of driver fatigue is a progressive dynamic process.Relevant research based on hidden Markov model(HMM)must determine the model′s initial values firstly and the training process tends to fall into local optimum.Therefore,particle swarm optimization(PSO)algorithm is introduced into the process of training HMM to improve the above existing problems.What′s more,the improved method and forward-backward(BW)algorithm are compared in details based on typical driver fatigue data set.Experimental and analytical test results show that the improved method is more accurate and stable than BW algorithm in driver fatigue prediction.
出处
《大连理工大学学报》
EI
CAS
CSCD
北大核心
2018年第2期194-201,共8页
Journal of Dalian University of Technology
基金
国家自然科学基金资助项目(51675077)
中国博士后科学基金资助项目(2015M581329
2017T100178)
中央高校基本科研业务费专项资金资助项目(DUT16QY42)
关键词
驾驶疲劳
隐马尔可夫模型
前向后向算法
粒子群优化算法
driver fatigue
hidden Markov model
forward-backward algorithm
particle swarm optimization algorithm
