BCI System Of Spontaneous EEG Enhancement Based On Multi-modal Training

Brain-Computer Interface(BCI)is an emerging technology produced by the development of artificial intelligence and neuroscience.It converts the neuronal activity signals of the biological brain into instructions for controlling computers or other external devices.This technology is available used in medical rehabilitation,intelligent control and military training.Motor Imagery(MI)Electroencephalogram(EEG)is a type of spontaneous brain electrical signal,which refers to the brain electrical signal generated by imagining the movement of the limbs without actual physical movement.The brain-computer interface for motor imaging(MI-BCI)is the direct use of the brain's subjective thinking activity signals to achieve communication with the outside world or control of the surrounding environment.It is currently a hotspot of artificial intelligence control research.Among them,signal acquisition and processing are very important functional parts of MI-BCI system.However,the low signal-to-noise ratio,low acquisition rate,and individual variability of motor imagery EEG signals seriously limit the performance of MI-BCI system.In this paper,we start from enhancing the acquisition efficiency,feature information,and classification performance of motor imagery EEG signals combining EEG time-frequency domain-space domain features and multi-level EEG feature extraction module with synchronous coupling features to investigate how to most efficiently guide subjects in limb-motor imagery and fully activate the energy in the sensorimotor area of the subjects.Finally,we will achieve multi-dimensional improvement of the quality of MI EEG signal,and then improve the performance of MI-BCI system.The innovative results of the research are as follows:(1)A BCI system was designed and built for spontaneous EEG enhancement training based on multi-mode training and multi-level feature extraction.The system includes different imagination enhancement training module,multi-channel spontaneous EEG screening module,multi-level feature extraction module,feature downscaling module and intention recognition module.Firstly,different imagination enhancement training paradigms are designed including action execution training mode,visual imagination training mode and multi-modality enhancement training mode based on the information of spontaneous EEG generation characteristics,brain feedback mechanism and acquisition mode;secondly,the multi-modality enhancement training system is designed based on C#language that can realize the functions of signal acquisition,multi-scene training and data management.(2)A multi-channel spontaneous EEG signal screening strategy was proposed.Concrete can be described as: firstly,the spontaneous EEG signals of multiple people under different imagery enhancement training modules are collected for multiple days,and the data are independent among subjects,training methods,and training days.What's more,the patterns of current instructions and energy fluctuations in brain regions during motor imagery are analyzed based on brain topography maps to study the range of brain region enhancement after training.And then,the EEG channels with significant energy enhancement after training are screened from the enhancement regions to improve the speed and quality of signal analysis.Finally,the optimal time-frequency band of spontaneous EEG is screened by using continuous wavelet transform to analyze the temporal-frequency enhancement phenomenon of training.(3)Verify the effectiveness of the multi-modal training spontaneous EEG enhancement system and multi-channel signal screening strategy designed in this paper.On the one hand,the multi-level features of spontaneous EEG in the time-domain,frequency-domain,air-domain,and synchronous coupling under different imagined enhancement training are extracted to compare and analyze the enhancement phenomena at each feature level and finally verify the effectiveness of multi-modal training in improving spontaneous EEG signals.On the other hand,the multi-level features are optimally screened to analyze the changes of intention recognition rate for subjects under different classifiers,and to determine the optimal training system for spontaneous EEG signals by considering the influences of multiple factors such as individual differences of subjects and the stimulation effects of training modes.