Research On Fatigue Driving Detection Based On Physiological Electrical Signal And Facial Image

With the development of economy,there are more and more cars in China.However,with the increasing car ownership,the number of traffic accidents is also increasing,especially those caused by fatigue driving.Therefore,effective and accurate monitoring of the driver’s fatigue state in the process of driving can avoid or reduce the occurrence of traffic accidents.The research of fatigue driving detection based on EEG,ECG and other bioelectrical signals and driver face images has been widely concerned.However,the existing methods have some problems,such as the detection accuracy of fatigue driving needs to be improved and the model effect is unstable due to the individual differences of drivers.Therefore,this thesis proposes a fatigue driving detection model based on bioelectrical signal and facial image feature fusion and transfer learning,then verifies the performance of the model using public data sets.Firstly,this thesis introduces the related research of fatigue driving detection methods at home and abroad,analyzes the generation mechanism and influencing factors of driving fatigue,introduces the generation principle and characteristics of EEG and ECG,and analyzes the correlation between ECG,EEG,surface image and driving fatigue from the mechanism.Secondly,denoising EEG and ECG by wavelet analyze.Frequency domain features of EEG are obtained based on power spectrum,and heart rate variability features are obtained based on timefrequency domain analysis.CLM is used to mark the key points on the face and calculate the average distance between the upper and lower eyelids.At the same time,in order to reduce the change of the average distance between the upper and lower eyelids caused by the change of the face orientation and the distance between the face and the camera,the average distance of upper and lower eyelids is adjusted dynamically,and then extract PERCLOS as feature.Finally,aiming at the problem that the accuracy of fatigue driving detection model is affected by individual differences,this thesis proposes a fatigue driving detection model based on multi-source domain selection and transfer learning.The optimal source domain set is obtained by source domain selection,and the transfer learning is carried out for each source domain in the optimal source domain set,and the weighted voting is carried out for the output of each source domain model to output the final result.The accuracy and robustness of the model are verified by experiments.