Collaborative Brain-Computer Interface Based On Rhythmic EEG

Electroencephalography(EEG)-based brain-computer interfaces(BCIs)have been studied and improved since the 1970 s.At present,the main development field of BCI is still mainly clinical application,aiming to provide effective ways and methods for brain-injured patients to communicate with the outside world,reduce patients’ dependence on others,improve the quality of life of patients and users,and reduce the burden on medical staff.Due to factors such as large individual differences in single-person EEG signals and being easily affected by personal status,the information transmission rate of BCI is affected.In order to solve the problem of low ITR of single-user BCIs in brain-computer interface,this thesis proposes a different previous collaborative paradigms improved overall BCI performance by fusing multi-lead EEG signals from multiple users.At present,multi-person cooperation BCIs is still in the early stage of development,and has great development space and prospects.Rhythmic EEG signals can be collected non-invasively,and have high-quality frequency domain features for EEG signal classification.A collaborative brain-computer interface based on rhythmic EEG signals can effectively improve the overall quality of EEG signals by fusing the features of multi-user EEG signals.This thesis learns and trains a suitable EEG classification model for a collaborative brain-computer interface system through deep learning.Based on the characteristics of EEGnet,this thesis analyzes and improves EEGnet to adapt to the characteristics of EEG data collected in this laboratory,and designs a single-person convolutional neural network classification model.A rapid use of transfer learning methods to build laboratory classification models for the field of brain-computer interfaces is studied.This thesis uses the software environment of tensorflow2.0-gpu+python3 to construct the offline and online system of rhythmic collaborative EEG interface.The EEG signals from 10 volunteers were collected for research and experiments.Realize the effective combination of online data collection and offline classification model through socket.Realize the effective combination of online data collection and offline classification model through socket.Using a programmable robotic arm as a control unit to build a complete collaborative brain-computer interface online system.The accuracy of the online experiment matches that of the offline study.