Driving Fatigue Detection Model Based On Multi-entropy Feature Of EEG Signals

With the development of technology and economy,the variety of vehicles is increasing day by day,and the traffic accidents are increasing year by year.In various malignant traffic accidents,the probability of traffic accidents caused by fatigue is 3-5times higher than that of normal driving.Therefore,if EEG can be used to detect fatigue state in real time,it is of great significance to avoid traffic accidents.In order to accurately detect the fatigue state of the driver,this paper mainly studies the fatigue detection model under the multi-entropy feature,and analyzes and studies the three main steps of denoising,feature extraction and classification in the fatigue detection process to establish three accurate Fatigue detection analysis model with high rate and stable characteristics:(1)For the preprocessing of EEG noise reduction,a fatigue driving detection application based on wavelet transform and multiple indexes is proposed.Based on the wavelet transform noise reduction processing,the kernel principal component analysis(KPCA)method is applied.The feature extraction is carried out,and finally the least squares classifier is used for classification.The experiment proves that the data classification effect after noise reduction has obvious advantages compared with the data classification result without noise reduction processing.classification effect after noise reduction has obvious advantages compared with the data classification result without noise reduction processing.(2)Aiming at the multi-features of fusion EEG,a method of fatigue EEG classification based on integrated multi-feature fusion and three different kernel classifiers is proposed.The 32-lead EEG is preprocessed,and the abnormal mode and high-frequency noise are removed by using Empirical Mode Decomposition(EMD).Then,the sample results are extracted by sample entropy and fuzzy entropy.The feature set is taken as the sample input.In the adaptive boostint classifier integrating three different cores,the experimental results show that the multi-feature extraction classification results are significantly higher than the single feature extraction results based on the same noise reduction processing results and classifier selection.(3)Aiming at the singularity of the classifier and the complementary advantages of each classifier,a combined experimental method based on AdaBoost is proposed to analyze the EEG detection fatigue driving.During the experiment,different subjects were analyzed by Independent Component Correlation Algorithm(ICA),and then sample entropy,information entropy,fuzzy entropy and AR coefficients were extracted.Finally,AdaBoost was used to base the least squares vector machine.The three core classifiers are integrated into one strong classifier.The test results show that the classification effect of AdaBoost classifier is better than that of single nuclear classifier.The average recognition rate of fatigue driving is 93%,and the accuracy rate of five-foldcross-validation is 91.04%,which broadens the safe driving based on EEG signals to some extent.The research path of the auxiliary monitoring system.