Research On Driving Fatigue Detection Method Based On Multi-physiological Information Fusion

With the rapid development of the global automotive industry,automobiles have gradually became a necessity in people's daily lives.The popularity of family cars has brought great convenience and enjoyment to people in both work and life,but it has also increased the rate of traffic accidents.According to the statistics,traffic accidents caused by driving fatigue account for more than 53% of the total number of traffic accidents in the world and are increasing year by year.It can be seen that effective monitoring of the driving fatigue has an importantly practical significance for reducing or avoiding the traffic accidents.Large studies have shown that the analysis of the physiological signal is the most objective and effective way to monitor the fatigue in humans.For this,this thesis establishes the driving platform for simulation to simultaneously acquire the EEG,EMG and ECG data during driving.Based on the dynamic feature extraction method,the muti-dimension features in both time-domain and frequency-domain are extracted for EEG,EMG and ECG signals,respectively.After that,the features are fused and the pattern recognition are conducted to identify the fatigue status.These strategies contribute to the effective monition and evaluation of driver fatigue status from the view of the mental,behavioral and psychological.This research provides a novel approach to analyze the driving fatigue based on the fusion of multivariate physiological signal.The main work of this thesis includes:Firstly,set up the driving platform for simulation to simultaneously acquire the physiological signals of the driver's brain power,myoelectricity,and electrocardiogram under different fatigue states during the simulated driving process,and at the same time subjective questionnaire investigation and analysis of drivers' driving fatigue.Secondly,preprocessing the EEG signals obtained from the experiment,analyzing the changes of EEG signals at different frequency bands during the fatigue generation process,determining the calculation method of EEG signal fatigue indicators,and statistically analyzing the trends of the EEG fatigue indicators as driving time increases.,And to prove the accuracy of the selected EEG fatigue index for fatigue response.Thirdly,imultaneously pre-process the EMG signal and ECG signal during the experiment,analyze the intermuscular consistency and mean instantaneous frequency of the EMG signal,and the heart rate and heart rate variability of the ECG signal.Study the trend of each characteristic index of electromyography and ECG signals with driving fatigue.Finally,carry out multi-physiological signal feature fusion and dimensionality reduction,and use the support vector machine classifier to detect driver fatigue.At the same time,in order to improve the accuracy of recognition,a driving fatigue detection method based on depth belief network is proposed,and a deep belief network model is used.Multi-physiological signals were used for feature fusion and fatigue state detection,and the fatigue detection results of the two classification models were compared and analyzed.