基于生理参数的驾驶疲劳检测方法综述  

A Review of Physiological Signals-Based Driving Fatigue Detection Algorithms

作　　者：苏瑞芝 唐巾卜 阿地力·吐合提 罗竞春 陈晨 陈炜 SU Ruizhi;TANG Jinbu;ADILI TuHeTi;LUO Jingchun;CHEN Chen;CHEN Wei(School of Information Science and Technology,Fudan University,Shanghai 200438,China;Human Phenome Institute,Fudan University,Shanghai 201203,China)

机构地区：[1]复旦大学信息科学与工程学院,上海200438 [2]复旦大学人类表型组研究院,上海201203

出　　处：《复旦学报（自然科学版）》2023年第4期419-427,共9页Journal of Fudan University：Natural Science

摘　　要：疲劳会影响驾驶人员的注意、思维、判断、决定等,是导致交通死亡事故的重要因素之一。实现高精度、强鲁棒性、轻量化的驾驶疲劳检测算法对于提升公共交通安全具有重要意义。该综述从信号模态(单模态、多模态)及分析方法(机器学习、深度学习、迁移学习)两个角度回顾并总结了基于脑电信号(EEG)、眼电信号(EOG)、心电信号(ECG)和肌电信号(EMG)进行驾驶疲劳检测的最新进展。同时,该综述分析了现有工作的局限性并探讨了未来的研究方向,为进一步提升基于生理参数的驾驶疲劳检测算法性能和推广结合可穿戴设备的实际应用提供了新颖的见解。Fatigue has an important impact on driver s attention,thinking,judgment,decision-making,etc.,and it also contributes as one of the leading causes of death among the traffic accidents.For public transport safety,high-precision,robust and light-weighted algorithms for driving fatigue detection are of great significance.This paper presents a review of driving fatigue detection technique based on physiological signals,including electroencephalography(EEG),electrooculography(EOG),electrocardiography(ECG),electromyography(EMG)from two perspectives:signal modality(single-modality,multi-modality)and analysis method(machine learning,deep learning,transfer learning).Meanwhile,the advantages/disadvantages and the future directions of the existing studies are discussed.The driving fatigue detection algorithms generally extract features through handcrafted features or automatic features extraction algorithm,combined with artificial intelligence algorithm for detection.The temporal relationship and spatial relationship between samples also play a role for improving the performance of the algorithm.Using multimodal signals can improve the robustness and accuracy of the algorithm.In order to better apply to practical application scenarios,the transfer learning algorithms can be used to reduce the domain shifting and improve the generalization ability of the algorithm.This paper also provides potential insights to improve performance and promote possible applications by integrating the algorithms with wearable devices.The future research can focus on the artifacts removal,channels reduction,modal decomposition,algorithm lightweight,and domain generalization.

关 键 词：驾驶疲劳 脑电信号 眼电信号 心电信号 肌电信号 

分 类 号：U471.1[机械工程—车辆工程]