Study On The Driving Fatigue Assessment Method Based On Multi-physiological Signals And Transfer Learning

With the rapid development of society,people's living standards continue to improve,and cars have become an important means of transportation in daily travel.However,while the car gives people great convenience and enjoyment,it also increases the incidence of traffic accidents,especially the traffic accidents caused by driving fatigue.It can be seen that effective monitoring of the fatigue state of the driver during driving has important practical significance for reducing or avoiding the occurrence of traffic accidents.At present,driving fatigue assessment methods based on multiple physiological signals have received extensive attention.However,the accuracy of driving fatigue assessment needs to be improved and affected by individual differences.Therefore,this paper proposed a driving fatigue assessment method based on EEG,ECG and EMG feature fusion and transfer learning,and applied the method to the driving fatigue assessment system.The specific work of this paper is as follows:Firstly,This paper introduced the current research status and development trend of assessment methods for driving fatigue at home and abroad,and then proposed the research content based on multi-physiological information fusion and transfer learning.The mechanism and influencing factors of driving fatigue were analyzed.The principle and characteristics of physiological signal generation were introduced,and the relationship between physiological signal and fatigue was studied from the mechanism.Secondly,this paper synchronously collected EEG,ECG and EMG signals during the driving process of 20 drivers,based on the simulated driving platform design experimental paradigm.The artifact interference of each physiological signal was introduced,and the artifact noise interference was removed by filter and independent component analysis.The EEG nonlinear fatigue index was extracted by the wavelet sample entropy.The EMG frequency domain index and nonlinear index were extracted by the power spectrum and approximate entropy.The heart rate and heart rate variability of the ECG were extracted by the time-frequency analysis method.The fatigue indices were feature fusion.Thirdly,according to the difference of fatigue index distribution of different drivers' physiological signals,a driving fatigue assessment model based on the fusion of transfer learning and decision-making level was proposed.The initial judgment of the fatigue state from a single source domain to the target domain was implemented based on the Large Margin Projected(LMPROJ)Transductive Support Vector Machine.Then,the evaluation results of the above multiple single source domains to the target domain fatigue state were combined at the decision level to improve the accuracy and robustness of the driving fatigue assessment.Finally,the driving evaluation system based on multi-physiological information and transfer learning was designed,including the design and implementation of hardware and software functions of the fatigue assessment system.The driving fatigue assessment method proposed in this paper was applied to the system.In order to verify the superiority of the method,the paper designed a comparative experiment.The experimental results showed that the driving fatigue assessment method based on transfer learning had higher accuracy and robustness.