EEG Signal Analysis Based On Riemann Space And Its Application In Rehabilitation Brain-computer Interface

With the progress of science and technology and the continuous development of the times,the research and application of brain-computer interface(BCI)has attracted more and more attention,and has become a global frontier technology development trend.BCI refers to the establishment of a new information exchange and control channel between the brain and the external environment without relying on the conventional spinal cord / peripheral neuromuscular system,so as to realize the direct interaction between the brain and the external equipment.In other words,BCI enables external devices to read the brain nerve signal,and converts thinking activities into command signals to realize the manipulation of human brain thinking on external devices.It has important application prospects in brain science,rehabilitation medicine and military fields.Motor imagery based BCI is a kind of spontaneous BCI system.It does not need external stimuli(such as stroboscopic,etc.),and can generate EEG signals only through brain imagery.This is also closer to the original intention of BCI research,that is,mind control.However,due to the weakness and low signal-to-noise ratio of the EEG signals,there are few effective features that can be extracted from the motor image EEG signals.As a result,the application of the motor imagery-based BCI is greatly limited.In a BCI system,the signal analysis method is the key point.At present,the traditional EEG signal analysis methods generally exist in Euclidean space.However,based on the research,the characteristic description and characteristic differences of EEG signals in Euclidean space cannot meet the requirements of BCI system.In recent years,the Riemannian space analysis method has attracted the attention of many researchers.The Riemannian space method can directly operate the space and its subspace composed of signals on the basis of its geometric structure.Compared with the traditional Euclidean space processing method,it has more superior information extraction and analysis performance.In view of the existing problems,this thesis proposes an EEG signal analysis methd based on Riemannian space,which proceses and analysis the EEG signals in Riemannian space.The correspoinding off-line and on-line experiments are conducted to verify the efficiency and effectiveness of the proposed method.The main research contents include the following parts:(1)The research status of EEG signal processing at home and abroad is introduced,and the advantages and challenges in the development of EEG signal classification in Riemannian space are discussed.(2)The overall system architectureof the EEG analysis method in Riemannian space is designed.Aiming at the key problems,corresponding solutions are proposed based on the indepth analysis of the characteristics of EEG signals and Riemannian space,including using the Power Spectral Density(PSD)matrix as EEG signals conversion operator from European space to Riemannian space,calculating two free points' Riemannian distance with the aid of fiber bundle theory,defining the Riemannian distance between two PSD matrices as the dissimilarity measure of two EEG signals,and using k-nearest neighbor algorithm to classify multichannel EEG signals.(3)The corresponding off-line and on-line experiments were designed and conducted to verify the efficiency and effectiveness of the proposed method.Experimental results show that with same data set,the accuracy of the proposed method is better than some other Riemannian algorithms and some traditional Euclidian space methods.The highest recognition accuracy of the proposed method can reach 86.67%.Besides,the proposed algorithm has higher efficiency and shorter computation time.Finally,the proposed method was tested on the rehabilitation robot system and the experimental results verify the practicability and effectiveness of the proposed method.