Research On The Recognition Method Of Multi-class Motor Imagination EEG Signal And Its Application In AR

With the development of science & technology and the demand of the times,humans’ cognition and understanding of the brain is getting deeper and deeper,brain science has gradually become a popular research field for researchers,as a direct application of brain science research,the brain-computer interface(BCI)has also received widespread attention.The BCI system based on motor imagery(MI)can bypass nerves and muscle tissues,and control external devices only by imagining the brain electrical signals generated when the limbs are moving.This has very important application value and practical significance for helping people with normal brain activity but nerve damage or physical disabilities.This paper has conducted an in-depth study on the recognition method of MI electroencephalogram(EEG)signals.According to the characteristics and generation mechanism of EEG signals,a recognition method based on VMD-CSP and GWO-TWSVM is proposed.At the same time,developed an AR roaming system driven by EEG signals,using augmented reality(AR)technology.Completed the virtual character roaming control experiment based on MI,which provides theoretical support for the establishment of a fast,reliable and stable online MI BCI system.The main research contents of this paper are as follows:(1)Research on feature extraction method of MI EEG signals.Aiming at the three-class MI EEG signals of the shoulder collected in data set 1 of this paper,a feature extraction method based on VMD-CSP is proposed.The collected MI EEG signals are decomposed using a variational mode decomposition(VMD)algorithm to obtain intrinsic mode functions with different bandwidths and center frequencies,and then these intrinsic mode functions are input into the common spatial pattern(CSP)to extract features,which validates the method effectiveness.(2)Research on classification method of MI EEG signals.A GWO-TWSVM algorithm is proposed to classify and recognize the three-class MI EEG signals of the shoulder collected in data set 1.In order to enhance the classification performance of the twin support vector machine(TWSVM),the grey wolf optimizer(GWO)algorithm is used to find the most suitable TWSVM parameters for each subject,and then the EEG signals when the subject performs different MI tasks are identified,and comparing with the other four commonly used recognition methods,it proves the distinction of the classification method proposed in this paper.(3)Research on brain-computer interface system based on MI.In order to confirm the utility of the theoretical study on this paper,an AR roaming system based on MI EEG signals was developed.The signal processing and transmission are realized through the hybrid programming of MATLAB and C#,the client interface is programming using Visual Studio 2017,and the AR scene and character model are built in Unity 3D.Control experiments were carried out using the subjects’ MI data collected in the data set 2 of this paper.The results showed that the control accuracy of all subjects was 85% and above.It proves the validity of the academic research of this paper and the feasibility of the AR roaming system based on MI EEG signals,which has certain practical value.