| Fatigue driving is one of the main causes of traffic accidents.Most drivers experience fatigue with reduced alertness during driving,which seriously endangers road traffic safety.Therefore,it is necessary to detect the driver fatigue effectively and quickly and take some measures.In recent years,the detection of driver fatigue has been widely concerned by scholars.The main purpose of this study is to find physiological indicators that can represent driver fatigue and propose corresponding detection algorithms,so as to realize the detection of fatigue start state.Since EEG signal has always been recognized as the "gold standard" signal for fatigue detection,EEG signal has high signal-to-noise ratio and is easy to collect,and PERCLOS is one of the accurate indicators for fatigue detection under ideal experimental conditions.So this paper studies the driver fatigue detection algorithm based on EEG,EEG and eye movement signals.Since the discovery of EEG signals,the research of fatigue detection based on EEG and EEG has made great progress.In the process of fatigue detection research,the appearance and disappearance of alpha wave in EEG signal is considered as a reliable sign of fatigue.Therefore,in the research process of this project,we will focus on investigating the change rules and characteristics of various waves including alpha waves when fatigue occurs,and explore whether they can be used as reliable physiological indicators of driver fatigue and propose corresponding detection algorithms to achieve automatic detection of these physiological indicators.The main contributions and innovations of this paper are as follows:(1)A simulated driving experiment is designed.Using the simulated cockpit to simulate fatigue driving,a total of 15 subjects were collected 1.5 hours of simulated driving test data per person.At the same time of inducing fatigue signals,wearable equipment is used to collect EEG and EEG signals,and eye movement instrument is used to mark data.(2)Explore the ability of deep neural network to rely on EEG data modeling.In order to better analyze the expression of fatigue in EEG signals,which is different from the feature vector expression method of traditional EEG analysis,this paper converts the EEG signals into an image signal that retains time,space and frequency,and inputs it into a recursive convolution neural network for classification.(3)The feasibility of the designed recursive convolution neural network fatigue detection system is verified on the common data set SEED-VIG.The SEED-VIG data set only collects a small number of EEG and EEG electrodes,in order to study the optimal fatigue channel,collect EEG and EEG signals as much as possible in the experiment,and lay a foundation for the development of practical wearable fatigue driving detection equipment based on EEG and EEG signals.(4)A fatigue detection system based on single-channel EEG signal is proposed to extract the EEG features of sample entropy,fuzzy entropy and differential entropy under fatigue and normal conditions.The classifier compares the gradient lifting decision tree,K-neighborhood and support vector machine,and finally achieves the best classification effect on the gradient lifting decision tree classification system based on differential entropy. |
PDF Full Text Request