Research Of Vigilance Estimation Model Based On Forehead Electrooculogram

Vigilance is usually defined as the sensitivity or ability of people tomaintain attention and alertness to external stimuli, when they perform sometasks during a long time. Vigilance is a physiological indicator closely relatedto the waking or fatigue state. People’s reaction time to external changes orresponding accuracy for a long period can be used in evaluation of the degreeof people’s alertness levels.When people feel fatigue or sleepiness, vigilance decreased significantly.For car and high-speed railway drivers, decreasing vigilance is extremelydangerous and would make increasing the probability of occurrence of trafficaccident greatly. Incidents caused by lower vigilance are considered as onekind of fatigue driving. Drivers’ fatigue as an important incentive for trafficaccidents has received widespread attention.Traditional vigilance analysis methods based on physiological signalsusually use EEG and EOG physiological characteristics. The electrodescontacted with the human body directly can offer more reliable signalscompared to video signals. On the other hand, however，the electrodes foracquiring EEG and EOG signals may lead inconvenience and hinder the usersin practical applications.In this study, to overcome the shortcomings of traditional electrode placement for acquiring EOG signal, we place the electrodes on the foreheadand use forehead EOG signals to analyze vigilance, instead of the traditionalEOG signals. The main idea of this study is to use independent componentanalysis to separate independent EOG components from forehead EOGsignals. Blink detection algorithms and SVM-based classification method forremoving EMG and EOG artifacts are used for select independent HEO andVEO signals from extracted independent signals. In the end, independenthorizontal and vertical EOG signals are used as input signals for our vigilanceestimation system based on EOG signals.This thesis has two main contributions to EOG-based vigilanceestimation. The first one is that we put forward and prove the assumption thatthe horizontal and vertical EOG signals can be extracted from the foreheadEOG signals. The second one is that we propose a methodology for usingforehead EOG signals for vigilance estimation, including how to select theelectrode placement strategies, how to separate independent componentanalysis, and how to perform feature extraction and classification.The conclusion of this study is that independent HEO and VEO signalscan be definitely separated from the forehead EOG signals, and can be usedfor vigilance estimation. Based on the results of this study, it is possible togreatly enhance the practicality of EOG based vigilance estimation system,and make it truly become part of the vigilance detection solutions used by thedrivers of cars and high-speed railway.