Research On Classification Of Motor Imagination EEG Signal Based On Generative Adversarial Network

Brain-Computer Interface(BCI)provides a way to directly interact with the environment and control external devices without relying on peripheral nerves and muscles.It can directly read the physiological electrical signals in the human brain and analyze its meaning.And convert it into a control signal to control external equipment.BCI has been widely studied and applied in the fields of disease recognition,neural repair,neurofeedback training,the use of brain waves to generate visual images,emotion classification,military,entertainment and other fields.The classification of Motor Imagery Electroencephalogram(MI-EEG)signals is one of the key technologies in the research of brain-computer interface.This article aims at improving the classification accuracy of MI-EEG signals,and combines the characteristics of MI-EEG signals to preprocess the EEG signals.Analyzing and researching the process of,feature extraction and classification,the main work is as follows:1.In the preprocessing stage of MI-EEG signals: This article uses band-pass filtering,common average reference and independent component analysis to remove signals that are irrelevant to EEG signals,thus laying the foundation for the feature extraction and classification of EEG signals.2.In the stage of feature analysis and feature extraction of MI-EEG signals: analysis of traditional time domain analysis,frequency domain analysis,time-frequency analysis,spatial analysis,and Convolutional Neural Network(CNN)The research results of feature extraction methods such as method show that the effect of CNN to extract features is due to other methods.On this basis,in view of the characteristics of fewer samples and fewer features of EEG signals,the use of CNN to classify EEG signals is poor,and a new CNN feature extraction method is proposed.Compared with traditional EEG feature extraction methods,A better feature extraction effect has been achieved.3.In the classification stage of MI-EEG signals: Based on the above research,a method of EEG signal classification based on the GANs is proposed.This method introduces the improved CNN feature extraction method into the GANs,and improves the discriminator of the GANs.The experimental results show that the classification accuracy of the improved GANs is 6.54% higher than that of the original,and the three classification accuracy is 96.97%,which is 3.52% higher than that of the traditional convolution EEG classification method.The classification accuracy is high,which has a high reference value for the classification of multi class MI-EEG signals.