Research On Pattern Enhancement And Recognition Technology For Active Brain-computer Interface

As an important research branch of human-computer interaction,the development and application of Brain Computer Interface(BCI),the direct interaction channel between human brain and external devices,are increasingly extensive,not limited to the medical rehabilitation field at the beginning of the development,but has now expanded to daily life entertainment and even military fields.At present,BCI systems based on EEG signals can be divided into active BCI,passive BCI and reactive BCI systems according to system architecture and functions.Among them,the active BCI has the advantages of being controlled by spontaneous EEG signals,strong subjectivity,separation from system stimulation equipment,simple system composition,etc.At the same time,there are inevitable individual differences problem and "BCI Illiteracy" problem in active spontaneous EEG.Focusing on the above issues and facing the active BCI,this paper conducts research from the perspectives of enhanced-mode EEG generation paradigm and EEG pattern recognition,mainly focusing on signal generation methods,supplemented by signal processing methods.The main research contents are as follows:(1)Design of combined enhancement paradigm: Based on the generation mechanism of spontaneous EEG signals,the pronunciation-motor imagination(PMI)BCI combining pronunciation imagination with action observation guided motor imagination is proposed,and the EEG signal acquisition system for multiple mental tasks is designed and implemented.(2)Verification of the combined enhancement paradigm: 15 subjects’ EEG data of multiple mental tasks were collected,and two methods,channel analysis based on Event-Related Spectral Perturbation and brain energy mapping analysis,were used to verify the feasibility of the proposed combined enhancement paradigm PMI-BCI experimental scheme.(3)Pattern recognition: EEG data feature extraction and separation are carried out based on the calculation of the covariance of EEG data.Firstly,the frequency selective Filter-bank Common Spatial Pattern algorithm is proposed.Secondly,the algorithm research based on Riemannian geometry is carried out.After the above methods are verified by the public dataset,they are used for the self-collected dataset containing 15 subjects.The experiment shows that compared with the traditional paradigm or algorithm,using the adaptive Riemannian geometry algorithm for PMI EEG data classification is more feasible and accurate.(4)Mixed reality mode enhancement and online system simplification: An enhanced PMI-BCI system integrating the mixed reality technology is further developed.The PMI-BCI training system is simplified by developing Holo Lens2head-mounted display,and the EEG of enhancement mode is obtained by using the characteristics of the mixed reality technology.The feasibility of the system scheme is verified through offline and online experiments.