| Fatigued driving is the leading cause of traffic accidents.Monitoring the driver’s condition and timely fatigue reminders can effectively reduce the occurrence of traffic accidents.Studies have shown that physiological signals such as ECG can directly reflect the driver’s state.However,driver state prediction accuracy based on physiological signals is not high because the physiological acquisition in the driving environment is easily interfered with by external environmental noises and driving actions.Therefore,it is of great significance to research the quality assessment and classification of dynamic physiological signals to improve the accuracy of driver state judgment.This paper carries out targeted research on the quality assessment of emotional,physiological signals for driver condition monitoring.This paper’s main research work includes:(1)A driving simulation platform for simultaneous acquisition of ECG and EEG signals was built.The signals of 5 volunteers for 10 hours were collected synchronously by designing a fatigue driving simulation experiment.The marking method was designed to achieve manual marking of the availability of the collected signals.Furthermore,combined with the fatigue state questionnaire and the EEG fatigue state criterion,the labeling of the driver state corresponding to the ECG signal was realized;(2)A multi-threshold ECG signal quality assessment scheme for resourceconstrained nodes is designed for driver wearable ECG application scenarios;(3)To improve the accuracy of multi-lead ECG signal quality assessment,a multi-classifier fusion ECG signal quality assessment algorithm was proposed;(4)The influence of compressed sensing on ECG signal quality and driver fatigue prediction accuracy is discussed.A redundant Gaussian dictionary for physiological signal data compression is proposed,and we presented a signal quality evaluation scheme for improving the prediction accuracy of driver fatigue state under data compression.The progress made in this paper includes :(1)Based on the principle of risk minimization,the fatigue status questionnaire and EEG indicators can effectively complete the labeling of the quality of physiological signals and the status of corresponding subjects in the simulated driving experiment;(2)The multi-threshold ECG signal quality assessment algorithm has the advantages of short calculation time and fast evaluation efficiency.The quality assessment accuracy of four typical wearable dynamic ECG signals reaches 99.12%,and the proposed scheme has a good signal quality evaluation effect in the real-time active ECG monitoring with limited resources.(3)In the MIT-BIH open data set test,the method based on the fusion of multiple classifiers was proposed,which effectively improved the accuracy rate of dynamic ECG signal quality assessment to 93.50%;(4)the redundancy of the Gaussian dictionary,with sparse binary measurement matrix,BPDN,and SL0 reconstruction algorithms,effectively improves the efficiency of compression perception reconstruction of ECG and pulse signal.It decreases the remote data transmission and the computational complexity to save resource consumption,to meet the distortion in the diagnosis of the premise,the most excellent compression of ECG and pulse signal rate can reach 70%;(5)Finally,we compared the scheme without signal quality assessment,the average accuracy,specificity,and driver state prediction scheme’s sensitivity.If it includes the physiological signal quality assessment steps proposed by us,reaching 86.24%,84.63%,and 84.03% for the wakefulness/fatigue state prediction of the tested object.This paper’s research results further enrich the research on the quality assessment of dynamic physiological signals,provide theoretical basis and technical support for collecting emotional,physiological signals,long-term continuous monitoring of driver status,early prevention of traffic accidents,and intelligent prediction from the technical level. |
