Detection System Of Driver Fatigue Based On Multi-biological Signals

Recent years, driver fatigue becomes one of the important causes of serioustraffic accidents. Therefore, the research in detection methods of driving fatigue hasbecome an important issue.Firstly, we set up a driving simulation platform, and design a strict experimentalparadigm to simulate the progress of driving fatigue. In the experiment, we introducevisual and auditory stimuli task. Through8steps of simulative driving mission, driverfatigue will be induced. After every step, subjects are asked to finish Karolinskasleepiness scale as a reference for the result of signal analysis. Also, visual andauditory stimuli task and driving result are recorded. Finally, the following severalaspects of the driver’s fatigue is studied in this paper.1) EEG signal is analyzed by the power spectral, microstates and EKG signal isanalyzed by heart rate. When driver get fatigue, the power spectralF()/wouldincrease andF/would decrease. And microstates index R would decrease whendriver hard to drive. The result shows that the power spectral and heart rate is closelyrelated to driver fatigue. We promote an algorithm to calculate heart rate. As a result,we verify the conclusion that driving fatigue cause driver’s heart rate decrease.2) In this paper, an optimized method is proposed to detect blink which based onAdaboost algorithm. Adaboost algorithm and Haar characteristic are used to train face,open-eye and close-eye classifier. According to face location, eye blink can be quicklydetected. Blink square wave figure and simplified PERCLOS feature are used to judgedriver fatigue. Result shows that the PERCLOS index will increase more than100%when driver is fatigue.3) Data fusion method is discussed to enhance the accuracy in judgment ofdriving fatigue. This paper adopts the method of support vector machine (SVM) andartificial neural network, to merge EEG and EKG on feature layer. The result shows,both SVM and artificial neural network can enhance the accuracy.4) A detection system of driver fatigue based on multi-biological signals isdesigned and developed. EEG, EKG and Blink is used to detect fatigue by this system.The problem of compatibility of multi-device is solved by this system, and designpattern is applied for data forwarding and processing. Experiment shows this system can judge fatigue or sober effectively.