Research On EEG Feature Analysis And Recognition Technology Of Brain Vigilance

Vigilance refers to the sensitivity of the operator to maintain long-term attention and vigilance when performing a certain task.In actual life,workers in many positions need to maintain long-term vigilance and the level of vigilance.The decline could affect people’s productivity and even cause safety incidents.Therefore,the objective detection of alertness level has important research significance and value.In this paper,a psychomotor Vigilance Test(PVT)and n-back are designed to construct a model of brain alertness reduction through a 3-back experiment,and then use PVT to detect the alertness during the entire experimental process.The experiment records the participants’ behavior and EEG data throughout the experiment,and fills in the subjective scale.The characteristic analysis of subjective scales,behavioral data,time domain,frequency domain,nonlinearity and brain network of EEG signals is used to explore the neural mechanism of the brain in different alertness states,and the alertness state of the brain is classified by pattern recognition.Subjective scale analysis showed that the scale scores after the experiment were significantly higher than those before the experiment.Behavioral data analysis showed that subjects’ mean reaction time increased significantly after the experiment.It shows that the experiment successfully induced the state of reduced alertness of the subjects,and a model of decreased alertness was constructed.Subjective scales were significantly correlated with reaction time,so data were annotated with reaction time.The EEG signal analysis showed that with the decrease of alertness,the frequency domain analysis showed that the average power value of EEG at 4-30 Hz increased significantly,indicating that the subjects needed to consume more energy to complete the experimental task when they were in a low-alert state.The three power ratios of θ/β,(θ+α)/β,(θ+α)/(α+β)increased significantly,the slow wave component of EEG increased,the fast wave component decreased,and the brain became more fatigued;in the nonlinear analysis section,the value of the entropy of the sample decreased,possibly due to a decrease in brain activity activity,inhibition of neurons,a decrease in the level of excitability of the brain;in the brain network analysis part,the average path length of nodes decreases,global efficiency increases,and node out-degree changes.Mainly in the frontal and temporal regions,the in-degree changes are mainly in the whole brain,the degree of connection of the brain network is weakened,and the efficiency of brain information transmission is slowed down.In the aspect of alertness level recognition,a classification model of alertness level detection and recognition is constructed by using support vector machine,K-order nearest neighbor and naive Bayes algorithm.After comparison,it is found that the support vector machine has better performance in vigilance recognition,and is selected as the optimal classifier.The classification effects of alertness under the time windows of 10 s,30s and 60 s were compared and analyzed,and 60 s was selected as the optimal time window.Using the Relief F algorithm and the common channel selection algorithm of adding weights,the top ten common effective channels were screened out,and the average accuracy of the three categories of alertness was 96.65%.of portable devices provides a theoretical basis.