A Practical Method Of Vigilance Detection Based On EEG

Vigilance level is affected by brain fatigue,which can lead to a decline in vigilance level and longer reaction time.For example,fatigue driving is likely to cause traffic accidents,so the detection of fatigue-induced vigilance level is extremely important for safe driving.At present,electroencephalography(EEG)is one of the most potential methods and has been widely used in various studies.The purpose of this thesis is to detect the vigilance level by analyzing the EEG signals collected by wearable EEG system.Firstly,by analyzing the sleep deprivation EEG data collected by a non-portable EEG acquisition system,it was found that fatigue could reduce the power of alpha wave when eyes were closed.The frequency-domain and time-domain features of EEG were extracted and a regression model was established by using a forward search strategy to select the features.According to the reaction time,EEG data were divided into alert and fatigue states.The relative powers of four frequency bands were extracted as features,and the artificial neural network(ANN)and support vector machine(SVM)classifiers were used for binary classification of alert and fatigue states respectively.With the shortening of data length,when the window length is 2s,the accuracy of classification of testing set by ANN and SVM is slightly decreased,which is 0.95 and 0.90,respectively.Then,according to the results of feature selection,a headband consisting of 8electrodes was designed,in which 6 electrodes were located in the forehead and 2electrodes were located in the occipital region.The designed headband was used for data recording with a wearable EEG system.The impedance,comfortableness,signal quality and classification accuracy of eye-open and eye-closed states for three kinds of electrodes(conductive fiber electrode,claw dry electrode and ECG plate electrode)were compared with data from a continuous eye-opening and eye-closing experiment.The results suggested to use the conductive fiber electrode in the frontal area and the claw dry electrode in the occipital region.Finally,a wearable EEG acquisition system was used to collect EEG data during sleep deprivation with the headband.It was shown that the alpha power in the fatigue state decreased by comparing with the alert state when eyes were closed.Because the channels were few and the noisy leads were not removed,not only the frequency domain features,but also the time domain features were extracted to build the binary classification model.By dividing the EEG data into alert and fatigue states,binary classification models for discriminating alert and fatigue states were established by random forest,SVM and ANN,respectively.The analysis results showed that the classification performance was relatively low when only using frequency domain features,and the accuracy was increased after the time domain features were added,leading to the highest accuracy of 0.92 when using ANN.In a word,this thesis studied EEG-based vigilance level detection algorithms and a scheme of wearable EEG acquisition,combining with feature extraction and pattern classification detection algorithm.Finally,a fast and effective vigilance level detection algorithm and a comfortable and convenient wearable EEG acquisition scheme are realized and demonstrated,which is of great significance for the practical applications of the EEG-based vigilance level detection technology.