Study On Driver Fatigue Detection Method Using Biological Electrical Signals

Traffic safety has always been one of the most severe problems in modern society. Oneof the significant reasons leading to critical accidents is driver fatigue driving, especially, inhighway traffic environments. The researches on the formation mechanism of drivers’fatigue driving and the states physiology, psychology and behavior and the detection of thestates of driving fatigue is highly meaningful to improve traffic safety and decrease trafficaccidents owing to drivers’fatigue driving.Today, the detection approaches of drivers’ fatigue driving mainly contain two kinds:the subjective detection approaches and the objective detection approaches. According tothe differences of the applied objective indicators, the objective detection approaches aredivided into three categories: the approaches based on vehicle state information, theapproaches based on drivers’ behaviors, and the approaches based on drivers’ physiologysignals. To the best knowledge of us, the detection of driving fatigue using human body’selectroencephalogram (EEG) is accepted as the most accurate and objective approach.A driver fatigue detection approach based on EEG is proposed with a review of researchand application status of driver fatigue approaches at home and abroad, such as the concepts,causes and influences of fatigue driving. With the introduction of EEG into the detectionapproaches of fatigue driving, the changing rules of EEG are analyzed and the evaluationstandards and detection approaches of the states of drivers’ fatigue according to the featuresof EEG are studies under the state of fatigue driving in the paperOn the basis of the research progress of driver behavior and bioelectricity in VehicleIntellectualization and Humanization Lab, human drivers in loop driving fatigueexperiments are designed and implemented using Driver In Loop (DIL) and BIOPACmultipurpose polygraph in this paper. In the experiments, the driver is driving for a long timeon a monotonous road. During the experiments, the changing process from sober to fatigue is experienced by every driver. The drivers’ EEG and subjective scoring of fatigue areconnected using drivers’subjective evaluation.The connected EEG signals are dealt using down-conversion resample, band-passfiltering etc. before feature extraction as the signals cannot be directly applied for the inertialcharacteristics of EEG. In this paper, the domain signals of four classic rhythm (、θ、、β) of EEG are extracted via the transform and reconstruction of wavelet. Then, the averagepower in the time window T=1min and the ratios of EEG rhythms are obtained by solvingpower spectrum. The ratios of average powerF1(α+θ)/βandF2θ/βare selectedas the character indicators of evaluating the ranges of driving fatigue.Using the changing rules of subjective evaluation and EEG characteristics, the drivingstates are divided into four grades, namely, sober, mild fatigue, moderate fatigue and severefatigue. The states of the driving fatigue are classified using SVM (Support Vector Machine),which can be utilized to solve the multi-class problems. A two-dimensional feature vectorincluding fatigue state feature indicators F1and F2is constructed and utilized to classify andidentify the four grades of the states of fatigue driving. The experiment data is divided intotwo groups: a training set and a testing set. Combined the scores of subjective evaluation,the statistics of the testing results verify that the proposed the extraction of fatigue statefeatures and SVM classification approach are valid and the states of fatigue driving ofdifferent extents can be detected using the approach using drivers’EEG signals effectively.