Study On Driving Fatigue Detection And Warning Method Based On Drivers’ Physiological Signals

Driving fatigue is one of the most important causes of road traffic accidents.It is of great significance to detect and warn the driver’s fatigue driving state.Based on the in-depth analysis of the current domestic and foreign research on driving fatigue detection and warning,the paper applies the photoplethysmographic(PPG)signal to the research of driving fatigue detection,analyzes the change characteristics of the PPG signal under the drivers’ fatigue state,and proposes a driving fatigue detection and warning algorithm based on support vector machine(SVM).Firstly,the paper uses the driver-in-the-loop test platform and pulse oximeter to design and implement a fatigue driving simulation experiment,and obtains the subjective evaluation fatigue level and PPG signals of the test driver from the awake state to the fatigue state when driving for a long time in the highway scene.The PPG signal and other data provide data support for the subsequent analysis of drivers’ fatigue status and the study of driving fatigue warning algorithm.Secondly,the PPG signal is weak,so the collected original signal is preprocessed by signal denoising and baseline drift removal before feature extraction.In the paper,the differential threshold method is used to identify the characteristic points,and then the relative amplitude of the main wave,the normalized amplitude of the dicrotic notch,the normalized amplitude of the dicrotic wave,the cardiac cycle,the ratio of rapid ejection period and the ratio of left ventricular ejection period are extracted as characteristic indicators.Through research,it is found that with the deepening of fatigue,most of the characteristic indicators have a relatively obvious trend of change,and there are significant differences between different fatigue levels,indicating that the selected characteristic indicators can be used to determine the state of the driver.Then,using the kernel principal component analysis(KPCA)method to reduce the dimensionality of the characteristic indicators,two principal components are extracted from the six characteristic indicators as new characteristic indicators,which can effectively reflect the drivers’ fatigue level.On this basis,the three sets of data samples are analyzed on classification.The result shows that the three sets of samples can be effectively separated by selecting the appropriate kernel function.Therefore,it is feasible to detect the fatigue state of the driver by the two principal components that are selected from the PPG signals.Meanwhile,the reduction in the number of characteristic indicators will reduce the training cost of the classifier and improve the classification speed of the classifier.Finally,a driving fatigue warning algorithm based on SVM is designed.The algorithm calculates the average value of the relative amplitude of the main wave,the normalized amplitude of the dicrotic notch,the normalized amplitude of the dicrotic wave,the cardiac cycle,the ratio of rapid ejection period and the ratio of left ventricular ejection period within 1 min of the detection period and then two principal components are calculated.The SVM classifier is used to determine the drivers’ fatigue level.If the drivers’ fatigue level is detected to be level two,there will be a weak warning.If the drivers’ fatigue level is detected to be level three,there will be a strong warning.The algorithm is realized using MATLAB and the experimental data is analyzed offline to complete the training and testing of the model.Finally,the driving fatigue warning algorithm program is written to the driver-in-the-loop experiment platform.By comparing the output results of the algorithm with the drivers’ subjective evaluation of fatigue level,the effectiveness of the driving fatigue warning algorithm based on SVM proposed in the paper is verified.