计算机科学 ›› 2015, Vol. 42 ›› Issue (3): 191-194.doi: 10.11896/j.issn.1002-137X.2015.03.039

• 人工智能 • 上一篇    下一篇

基于深度学习的疲劳状态识别算法

周 慧,周 良,丁秋林   

  1. 南京航空航天大学计算机科学与技术学院 南京210016,南京航空航天大学计算机科学与技术学院 南京210016,南京航空航天大学计算机科学与技术学院 南京210016
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受江苏省产学研联合创新资金项目(SBY201320423)资助

Fatigue Recognition Algorithm Based on Deep Learning

ZHOU Hui, ZHOU Liang and DING Qiu-lin   

  • Online:2018-11-14 Published:2018-11-14

摘要: 目前国内外的疲劳状态识别算法大多数是基于单一的、人为制定的疲劳状态特征实现的,且大部分算法结构复杂,效率比较低下,对驾驶员的个人行为习惯的适应性不强。为此,提出一种基于深度学习的疲劳状态识别算法,它引入深信度网络(DBN)来模拟输入图像数据分布,完成对疲劳特征的分层自动抽取,进而基于时间窗实现视频流图像的疲劳状态识别;同时,算法自适应调整网络学习率以减少网络预训练时间,采用反馈机制实现网络自进化,从而提高对用户个性化疲劳特征的适应性。实验结果表明,算法可以使网络获取到良好的疲劳特征,且误判率会随使用时间的增加而逐渐降低。

关键词: 疲劳状态识别,深度学习,深信度网络,疲劳特征,反馈机制

Abstract: Current domestic and overseas fatigue recognition algorithms are implemented using fatigue features which are mostly singular and man-made.Most of those algorithms have complex structure,low efficiency and weak adaptability for drivers’ individual behavior habit.To this end,the paper put forward a fatigue recognition algorithm based on deep learning.It introduces deep belief network (DBN) to simulate the data distribution of input images,extracts fatigue features automatically layer by layer,and then recognizes state of fatigue from video images based on time window.The algorithm adjusts the learning rate of the net adaptively to reduce pre-training time,uses feedback mechanism to let the net evolve by itself and as a consequence improves its adaptability for user personalized fatigue features.The experimental result shows that our algorithm acquires good fatigue features,and its misjudgment rate reduces gradually along with incremental time.

Key words: Fatigue recognition,Deep learning,DBN,Fatigue feature,Feedback mechanism

[1] Correa A G,Orosco L,Laciar E.Automatic detection of drowsi-ness in EEG records based on multimodal analysis[J].Medical Engineering & Physics,2014,6(2):244-249
[2] Li G,Chung W Y.Detection of Driver Drowsiness Using Wavelet Analysis of Heart Rate Variability and a Support Vector Machine Classifier[J].Sensors,2013,3(12):16494-16511
[3] 李伟,何其昌,范秀敏.基于汽车操纵信号的驾驶人疲劳状态监测[J].上海交通大学学报,2010,4(2):292-295
[4] 张希波,成波,冯睿嘉.基于转向盘操作的驾驶人疲劳状态实时检测方法[J].清华大学学报:自然科学版,2010,0(7):1072-1076
[5] 李绍文,王江波.驾驶员疲劳检测系统研究[J].计算机工程与应用,2013,9(15):253-258
[6] Hinton G,Salakhutdinov R.Reducing the Dimensionality of Data with Neural Networks[J].Science,2006,3(5786):504-507
[7] 余凯,贾磊,陈雨强,等.深度学习的昨天、今天和明天[J].计算机研究与发展,2013,0(9):1799-1804
[8] Hinton G E,Osindero S.A Fast Learning Algorithm for Deep Belief Nets[J].Neural Computation,2006,8:1527-1554
[9] 刘建伟,刘媛,罗雄麟.玻尔兹曼机研究进展[J].计算机研究与发展,2014,1(1):1-16
[10] Hinton G.Trainging Products of Experts by Minimizing Contrastive Divergence[J].Neural Computation,2002,4(8):1771-1800
[11] Neal R M,Hinton G E.A View of the EM Algorithm that Justifies Incremental,Sparse and other Variants[M]∥Learning in Graphical Models,1998,5-368
[12] 蒋斌,贾克斌,杨国胜.人脸表情识别的研究进展[J].计算机科学,2011,8(4):25-31
[13] Dasgupta A,George A,Happy SL,et al.A Vision-Based System for Monitoring the Loss of Attention in Automotive Drivers[J].IEEE Transactions on Intelligent Transportation Systems,2013,4(4):1825-1838
[14] 张伟,夏利民,罗大庸.基于人脸运动信息和改进保局投影的疲劳识别[J].计算机科学,2010,7(11):265-267
[15] Cyganek B,Gruszczynski S.Hybrid computer vision system for drivers’ eye recognition and fatigue monitoring[J].Neural computation,2014,6(SI):78-94
[16] 孙树亮,林雪云.基于记忆的SVM相关反馈算法[J].计算机科学,2011,8(10):256-258
[17] 陈云华,余永权,张灵,等.基于面部视觉特征的精神疲劳可拓辨识模型[J].计算机科学,2013,0(2):284-288

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!