Research On Warning Technology Of Driver Fatigue Based On Multi-source Information Fusion

With the development of transportation and increase of vehicle ownership in our country,traffic accidents have increased at an alarming rate, which is detrimental for an individual’ssafety and daily life. According to the latest research data, one of the main reasons for trafficaccidents is fatigue driving. As a result, monitoring the fatigue state of drivers accurately andproviding early warning at the same time plays an important role for avoiding fatigue driving,which has great implications in reducing traffic accidents. This research analyzed monitoringmethods of fatigue driving and established a driver fatigue experimental system which is basedon human-vehicle-driving environment. In the meanwhile, a mathematical model and evaluationrules of driver’s fatigue were established based on the method of fatigue driving informationfusion through D-S evidence theory and fuzzy logic theory.EEG is considered as the “golden criteria” of fatigue. Correlation analysis were conductedbetween EEG and respiration, EEG and pulse, EEG and heart rate to verify that respiration, pulse,heart rate are important signals to determine whether the driver is fatigue or not. Fatigueexperiments were conducted to collect respiration signal, pulse, heart rate and EEG of the drivers.Wavelet decomposition method was used to reduce noises coming from physiological signalsduring the experiments. Also, the same method was deployed to reduce the interference causedby the noise from baseline drift, body’s move and industrial frequency. Because the signalscollected by various sensors were different, different methods were used to extract the signalfeatures of respiration, pulse, heart rate and EEG. The results showed that there is a statisticallysignificant correlation between the signal of respiration, pulse and heart rate have and the degreeof fatigue.The warning method of fatigue driving based on D-S evidence theory and fuzzy logic theorywas proposed. Firstly, different data was collected when the driver was sober and under fatigue,respectively to establish subordinate degree function of fatigue samples. Then, a group ofunknown fatigue was tested to build the subordinate degree function. Subordinate degreefunction of fatigue data and the fatigue test data was matched. Where after, probabilityassignment function was established based on thesubordinate degree function, which facilitatedthe establishment of the decision rule. Finally, the decision rule was applied to determinewhetherthe driver was under fatigue.The research proposed test system of driver’s fatigue monitoring. Physiological signals ofthe driver were collected in two different places, which included indoor driving simulation testand outdoor road driving test. The results indicated that there is a consistency of a driver’sfatigue alteration between indoor and outdoor test. There was a marked change of the driver’spulse, heart rate, electronical activity of the brain, and respiratory physiology signal with the increase of the driver’s fatigue during driving.An early fatigue driving warning system which consisted of physiological informationcollection module, data processing module and early fatigue warning module was established.And the above methods were tested in the embedded early fatigue driving warning system. Thepulse, heart rate and EEG sensors of physiological information collection module can collect thedriver’s physiological information which then will be transported to the data processing moduleto be analyzed and evaluated. And a warning of fatigue will eventually be sent via the earlyfatigue warning module based on the evaluation from the prior module.