Study On Fatigue Identification Based On Wearable Electroencephalograph

In recent years,the development of wearable electroencephalograph provides opportunities for the development of fatigue identification and health management.The research aims to realize intelligent identification of fatigue and propose a fatigue identification mechanism based on wearable electroencephalogram(EEG).Firstly,the research conducted the EEG experiments to collect EEG data and corresponding fatigue degrees.Then,the research used Wilcoxon test,probit model,logist model,statistical analysis,support vector machine(SVM),random forests and other machine learning methods to explore the relationship between fatigue and its influencing fators,the relationship between fatigue and EEG,the fatigue identification model based on EEG data and the fatigue control with entertainment programme.Fatigue’s potential influencing factors included working hours,workload,working difficulty,sleeping duration,etc.These factors were potential causes of fatigue,while EEG are the expression of brain neuroelectricity of fatigue.The former were the causes,while the latter were the results.Understanding the causes and results of fatigue was an important prerequisite for the fatigue identification and health management.The Feature selection of EEG was to figure out which frequencies of the brain waves could present the fatigue.The results showed that α waves,β waves,θ waves and(α+θ)/ β,(β/α)were significantly different under fatigued and indefatigable conditions.These brain waves could help to present and identify different fatigue states.In the process of establishing the fatigue recognition models based on EEG data,the research applied and compared the algorithms of CART,C4.5,random forest and support vector machine,etc.The results showed that the fatigue random forest and fatigue support vector machine had better generalization ability,and the accuracy of data prediction was 83% and 88% respectively.In addition,the study considered that the data balance processing and data non-normalization processing could improve the fatigue prediction accuracy based on EEG data.Data normalization processing could easily lead to the mass loss of EEG data information,and should be avoided or processed in other ways.Finally,the research explored the fatigue difference before and after entertainment programme from the perspective of fatigue control.The study found that people over 35 years old were more likely to be relaxed by watching entertainment shows,while younger people under 35 years old had no significant improvement or deterioration in their fatigue before and after watching entertainment shows.