Analysis Of EEG Characteristics And Evaluation Of Operational Level In Human-Computer Interaction System

In modern society,with the continuous progress of science and technology,people use various technologies to improve the performance of the machine,so the machine becomes more and more complex,and the operator’s operation level is asked for higher and higher.How to evaluate the operator level more accurately has become a very meaningful research content.EEG signal is safe,convenient and noninvasive.In this thesis,EEG characteristics that can distinguish operators’ different operation levels are explored from the perspective of EEG signal.In order to evaluate the operator’s operation level with EEG characteristics,a human-computer interaction system with virtual inverted pendulum as the control object and joystick as the main controller is established in this experiment.Taking the operator’s operation level as the research factor,three stages of EEG acquisition experiments were set up to collect the EEG data of the same operator in three stages:primary stage,intermediate stage and advanced stage.The collected EEG data were analyzed by independent component analysis and preprocessing of adjust denoising algorithm,and the signal after removing the artifact was analyzed by power spectrum,recursive graph and sample entropy algorithm respectively,The EEG features which can distinguish different operation levels were extracted.The experimental results showed that the sample entropy of EEG data in the advanced stage was significantly smaller than that in the primary stage.The power spectrum changes obviously at different operation levels.In the recursive graph,we could see that the Dell and DET values in the primary stage were larger than those in the advanced stage,which fully showed that the complexity of EEG signals in the advanced stage was greater and the nonlinear characteristics were stronger.Support vector machine was used to classify the three EEG features.The accuracy of classification was more than 80%.After finding the optimal parameters by cross validation,the accuracy of classification was almost more than 90%.The classification effect was very good.The experiment showed that the three EEG features of power spectrum,sample entropy and recursive graph could clearly distinguish the operator’s operation level.