Research On Fatigue Driving Detection Method Based On Driver’s Facial Features And Physiological Signals

With the progress of society and the improvement of people’s living standards,cars have become an essential means of transportation in people’s daily lives.The accompanying traffic safety issues are also becoming increasingly serious,and traffic accidents have brought great losses to people’s lives and property.Fatigue driving is one of the main culprits causing traffic accidents.How to quickly and accurately detect the fatigue status of drivers and provide timely warnings has become an urgent problem that needs to be solved.This paper proposes a fatigue driving detection method based on driver’s facial features and physiological signals to address the issues of high false detection rate and poor stability of fatigue driving detection methods based on single features.This method integrates facial and heart rate features of drivers,effectively improving the accuracy of fatigue driving detection and reducing the limitations of single feature based methods.The main research content of this article includes:(1)Heart rate feature extraction.This article uses a non-invasive smartwatch to collect driver’s heart rate data,and then proposes a method for extracting heart rate variability features based on heart rate data by analyzing the characteristics of dispersed heart rate data.This method extracts six features,including Mean NN,SDNN,RMSSD,LF,HF,LF/HF,and extracts fuzzy entropy features of heart rate data based on an improved fuzzy entropy algorithm.(2)Facial feature extraction.Combining Multi Task Convolutional Neural Network(MTCNN)with Ensemble of Regression Trees(ERT)for face detection and localization.Use the MTCNN algorithm to annotate facial bounding boxes in facial images,locate 68 key points in the face using the ERT algorithm,and extract eye and mouth aspect ratio features based on the key points.(3)Establish the detection model.A fatigue driving detection method integrating facial features and heart rate features is proposed.The extracted facial features and heart rate features are fused at the feature level,and a fatigue driving detection model based on Long Short Term Memory Networks(LSTM)with the fusion feature sequence as the input is built.Divide the fused feature datasets into training,validation,and testing sets,train the model on the training set,optimize the model performance using the validation set,and finally test the detection model.Finally,experiments were designed to verify the superiority of the multi feature fusion method and the performance of the LSTM network model.The experimental results confirmed that the training accuracy of the model based on fused features is significantly higher than that of the model based on a single feature.The LSTM network model also has good performance under various evaluation standards,and can achieve more accurate fatigue state recognition.