Design And Implementation Of Brain-Computer Interacation System For Motor Imagery

The brain-computer interface is an interdisciplinary subject that has emerged in many fields in recent years.It is a new type of interaction method that does not depend on normal pathways such as peripheral nerves and muscles,and realizes direct communication between the brain and the computer.Motor imagery refers to imagining movement in the mind without having to make real movements.The motor imagery therapy is based on this.The patient performs repeated motor imagery and activates the corresponding brain regions,so that the nerve cells around the damaged nerve pathway Wake up and reconstruct the neural pathway to achieve the purpose of restoring part of the patient's motor function.The brain-computer interaction system for motor imagery studied in this paper is dedicated to assisting motor imagery therapy and promoting nerve rehabilitation.This paper completes the design and implementation of a brain-computer interaction system for motor imagery.Since the EEG signal is highly susceptible to interference during the acquisition process,the data should be pre-processed to remove artifacts before signal processing.In the following signal processing,the feature extraction of EEG signals is studied.Firstly,the common spatial pattern algorithm is studied.Then the most prominent cognitive phenomenon in the motor imagery signal is explored: ERD/ERS,based on this phenomenon put forward a paraments: ERD/ERS energy coefficient,which can be well divided into four types of motor imagery actions.Finally,the wavelet packet transform is introduced.A feature extraction method WPECSP is proposed,which extracts the approximate entropy of the node after decomposing the original data,and combines it with the CSP feature and the ERD/ERS energy coefficient feature.The experimental results show that compared with the CSP method,the classification accuracy of feature extraction using WPECSP is improved by 4.44%,reaching 97.22%.This proves the effectiveness of the method.In the model classification of EEG signals,the effects of Fisher linear decision and SVM are compared.When dealing with the contradiction between the two classifiers and the four classification tasks,the "one-to-one" and c-repetition points are compared.The effect of the two methods;when using the "one-to-one" method,the two methods of voting method and judgment decision function value are compared for the final output category judgment,and based on this,a voting method will be proposed.The method of judging the combination of decision functions,the experimental results show that the classification recognition rate is effectively improved.Finally,the influence of the time interval used in data processing on the experimental results is analyzed.According to the difference between the tested individuals,the optimal time interval is selected for classification,which effectively improves the classification accuracy.In the construction part of the brain-computer interaction system,based on Maya,the modeling and animation of the software robot was completed.Based on Unity3 D,the functions of the offline training system and the online subsystem were realized.The system can not only analyze the EEG signal offline,but also process it online and present the result in the form of the action of the software robot,giving the patient visual feedback and assisting the motor imagery therapy to improve the rehabilitation effect.