Multi-information Fusion Fatigue Driving State Detection System

The rapid development of motorization has increased the number of traffic-related casualties.Fatigue driving is the main cause of traffic accidents,and the public is still not aware of its potential hazards.Tired driving is called the "silent killer".Therefore,thorough research on the risk factors related to traffic accidents and fatigue-related casualties is essential.An estimate from the National Highway Traffic Safety Administration shows that the number of people who fall asleep while driving is increasing,and the resulting traffic accidents are countless.If there is an effective way to monitor this situation and alert the driver,many fatal accidents can be avoided.Many psychology researchers have discovered that when a person is drowsy,their heart signals change.In the early fatigue driving recognition,the effect of monitoring the driver's facial motion characteristics(such as frequent blinking and yawning)is not obvious,and monitoring their physiological changes is more efficient.Simple hardware settings are used to collect the driver's ECG(Electrocardiogram)data,and digital filters are used to remove noise and extract the required information for further analysis,which can make up for the lack of facial features of the driver in early fatigue driving recognition.When recognizing a severe fatigue state,monitoring the driver's facial features is more accurate and effective than monitoring the changes in heart rate features.Therefore,the fatigue driving recognition network model that integrates the driver's heart rate and facial information features is more accurate and satisfactory,and it can more effectively monitor the driver's fatigue at each stage.Multi-source information fusion to identify the driver's fatigue state not only greatly improves the accuracy of early fatigue state recognition,but also has an important role and practical significance to overcome the limitations of a single feature recognition method.(1)First,the paper discusses the method of identifying the driver's fatigue state;then,the design of the driver's fatigue state simulation experiment is carried out,and the comprehensive driver's self-evaluation and the fatigue state simulation experiment program of his evaluation are proposed,and the camera and ECG information collection are set up.The module-based simulation driving experiment platform obtains the driver's facial feature information,ECG feature information and subjective evaluation level information.The driver's multi-heart rate feature extraction method is based on the AD8332 ECG acquisition module to obtain the driver's 5-dimensional heart rate variability characteristics.(2)Using the driver facial feature information extraction method based on 68 facial feature points,from the driver's facial image information in different states,the driver's head posture and Euler angle,eye aspect ratio and mouth length-to-width ratio There are a total of 3 categories of 5-dimensional facial features;in order to avoid the influence of external light on facial image information,image preprocessing based on Enlighten GAN network is adopted;the influence of driver's head posture rotation on EAR and MAR is discussed,and Euler-based The angle correction makes the obtained eye aspect ratio and mouth aspect ratio more accurate.(3)The research on the principle and method of driver fatigue state recognition based on fusion of heart rate and facial features was carried out,and the real-time heart rate and facial features of the driver based on feature-level information fusion were studied,and an LSTM(Long Short Term Memory Network)with the fusion feature information as the input data source was constructed.The driver fatigue state recognition model.(4)A fusion feature data set fused with fatigue driving heart rate and facial information was created as the experimental data source,and the fatigue driving state recognition neural network model was trained and tested,and the recognition model was compared and analyzed from multiple dimensions.The results show that the method of fusing the heart rate and facial feature information to identify the fatigue driving state of the three levels of "normal state","fatigue state" and "severe fatigue state" three levels of fatigue driving state recognition accuracy rate is respectively 87.7%and 82.4 % And 85.5%,the average can reach 84.9%.