Design Of Brain-Computer Interface Stroke Rehabilitation System Based On Motor Imagery

Traditional treatment of stroke is achieved by indirectly stimulating the nervous system corresponding to the patient’s movements,which usually requires the involvement of the rehabilitation doctors.However,there is a large number of stroke patients and a shortage of rehabilitation doctors in some developing countries.Therefore,it is urgent to find methods to treat stroke patients that are not highly dependent on rehabilitation doctors.To this end,this thesis proposes a design idea of brain-computer interface stroke rehabilitation system based on motor imagery,which is tested offline on OpenViBE platform,including pre-processing,feature extraction and classification of Electroencephalograph(EEG)datasets.Then,the classification results are transmitted to the self-designed rehabilitation hand through serial communication.Nevertheless,the existing datasets used for offline testing can’t meet the EEG acquisition requirements for real-time rehabilitation,and the accuracy of the classification algorithm needs to be improved.To solve the above problems,this thesis improves the system and develops a set of EEG acquisition device and EEG analysis software.It is worth noting that the selfdesigned EEG acquisition system is successfully realized in this thesis,which plays an important role in promoting the development of EEG acquisition technology.The main contents are shown as follows:(1)A set of multi-channel EEG acquisition device is developed independently.Firstly,the design of EEG sensor array is completed,and two schemes of wet electrode and dry electrode are given.Secondly,the low noise and high sensitivity EEG acquisition circuit is designed,including analog front-end acquisition module,main control module and wireless communication module.Finally,the upper computer software with the functions of filtering,fast Fourier transform and impedance monitoring is designed.It should be mentioned that the developed EEG acquisition device can realize wireless transmission and multi-channel,high signal-to-noise ratio EEG acquisition.In addition,EEG analysis software with spectrum analysis,traceability analysis and feature information extraction is successfully developed.(2)A design idea of Brain-Computer Interface stroke rehabilitation system based on motor imagery is proposed,and verified offline based on the existing datasets of OpenViBE.Firstly,the datasets are preprocessed by Laplacian space filtering and Butterworth bandpass filtering.Secondly,the logarithmic band features are extracted from the datasets using the logarithmic band power,and the information features are classified by the linear discriminant analysis algorithm.Then,the hardware circuit of rehabilitation hand is designed by using Arduino microcontroller,relay air pump and air valve.Finally,python is used to realize the serial communication between OpenViBE and Arduino microcontroller,which successfully realizes the one-to-one correspondence between classification results and rehabilitation hand movements.(3)To improve the EEG signal acquisition and processing scheme,a new experimental paradigm is designed independently and different machine learning algorithms are used to analyze the collected datasets.Firstly,the experimental paradigm is written by Eprime to induce EEG signals by visual and auditory stimulation.Then,it is time to collect good quality twocategory EEG signals.At the same time,the synchronization box is used to mark the event time points to better process the EEG datasets.Secondly,after unpacking the collected EEG datasets,the finite length unit impulse response filter and infinite impulse response filter are used for preprocessing.Eventually,the processed datasets are applied to verify the feasibility of four different algorithms:logarithmic band power combined with linear discriminant analysis,common spatial pattern combined with linear discriminant analysis,common spatial pattern combined with support vector machine,and automated machine learning.It provides algorithm support for the subsequent design of real-time rehabilitation system.In addition,for quantitative analysis of closed loop neural circuits,a rehabilitation evaluation method is proposed.