Reasearch On Classification Of P300 EEG Signals

It has been the research direction of Brain-Computer Interface(BCI)technology to help patients with movement disorders and neurological diseases establish a way to communicate with others.This technology sets up an information channel between the experimenter brain and the computer device,transmits the EEG signals generated by the human brain and obtains the information expressed by the subjects.The visuallyevoked P300 is a common EEG signal,which has the advantages of uncomplicated experimental paradigm and no demand of special training for the subjects.By processing the P300 signal,the subject's imagine can be transformed and converted into actual work instructions,which has a great significance in the field of medical rehabilitation and industrial control.However,the signal-to-noise of the P300 EEG signal extracted by the non-invasive method is relatively low,usually mixed with other interference components,which affects the final classification accuracy.In view of the above problems,this paper takes the classification based on P300 EEG signals as the research content,combined with the relevant theoretical knowledge of deep learning,respectively to analyze and study the problems of denoising and recognition of P300 EEG signals.The main research contents and work of this article are as follows:(1)An parameter optimized adaptive denoising method based on variational mode decomposition(ADM-VMD)is proposed.The original EEG signal contains a variety of artifact signals and power frequency noise,which has a huge impact on the recognition of EEG signals.In this paper,the variational modal decomposition has been applied to segregate the original EEG signal into narrow-band sub-signals with different center frequencies.The signal components and noise components are divided for different sub-signals.Process and complete the reconstruction of the signal.At the same time,the standard firefly algorithm is used to optimize the parameters of the variational mode decomposition.The result of experiments indicates that this method improves the signal-to-noise ratio of the original signal to a certain extent.(2)Propose a P300 classification with attention mechanism(PR-AM)that integrates attention mechanism.In terms of EEG classification and recognition,the mainstream method is still based on traditional machine learning algorithms.Considering that these methods only use a single time-domain feature,the accuracy of classification is generally not high.In this paper,the time-domain feature and the spacedomain feature are fused to improve the classification effect of the network.At the same time,considering the human visual system's tendency to pay attention to image-assisted judgment information and the ignorance of irrelevant interference information,this paper embeds an attention mechanism in the feature extraction module to highlight key features related to P300 signal recognition.The result of experiments indicated that the PR-AM method has achieved excellent classification results in the BCI Competition dataset III 2004 dataset.(3)Design and implement a classification and recognition system based on P300 EEG signals.The system mainly preprocesses the collected P300 EEG signals,and then displays the information contained in it.The system can verify the practical effectiveness of the denoising method and classification method designed for the P300 EEG signal proposed in this paper,and can be applied to the actual experimental environment.According to the characteristics of P300 EEG signals,this paper formulates corresponding preprocessing methods and classification methods.From the results,the proposed methods have achieved certain results.Later expansion studies laid the foundation.