Research On Detecting Driver's Fatigue Based On Millimeter Wave Radar

At present,the traffic accident rate is high,among which 60 percent of the factors are caused by drivers,fatigue is one of the factors that cannot be ignored.If we can effectively detect the moment when the driver is tired and timely warn to remind,it will protect the safety of the person,the reduction of traffic accident rate is of great benefit.At present,the fatigue detection mainly focuses on three aspects: physiological signal,individual activity and vehicle behavior,among which the physiological signal detection results are the most accurate.However,electrocardiogram detection instruments are too numerous to be applicable to actual driving situations,and EEG and EMG detection cannot establish a general model due to the different physical characteristics of the subjects.Aiming at the deficiency of physiological signal in fatigue detection,this paper analyzes the driver's mental state in real time based on millimeter wave radar detection of physiological signal.Firstly,the experiment of non-contact fatigue driving was designed and carried out with the physiological signals collected by millimeter wave radar as the research object.Secondly,the collected original thoracic cavity signals were preprocessed,and two physiological signals,ECG and respiration,were separated by unwinding,bandpass filtering and frequency domain transformation.Then,the two physiological signals were studied in depth,the characteristic parameters derived from the analysis were extracted,the variation rule of the characteristic parameters with the deepening of driving fatigue was discussed,and the validity of the data was verified by statistical method and the accurate time of the occurrence of fatigue was determined.Finally,a suitable stochastic forest fatigue evaluation model is established to complete the core task of determining the time of driver fatigue.Seven core characteristic parameters(mean of heart rate,heart rate root mean square difference,heart rate low frequency,heart rate high frequency,ratio of heart rate low frequency to high frequency,mean of breathing,heart rate to breathing ratio)were used as the input of random forest,and the final overall accuracy was 91.67% after the test set classification test.The driving fatigue evaluation model trained by the random forest model has a high accuracy,which proves that the fusion of ECG and respiration characteristicparameters can effectively identify whether the driver is in the fatigue state.This method provides a new research idea and reference method for the detection of driving fatigue.The super-parameters of the random forest were optimized by Bayesian optimization,and the prediction accuracy of the optimized random forest model was 92.4%,which was 0.08%higher than that of the previous model.A new research idea and research direction were proposed for fatigue detection.