Research Of Motor Imagery Recognition And Proprioceptive Feedback On EEG Controlled Prosthesis

As one of the cleverest organs in human body, hands play an important role in theprocess to know the world and to Change the world. Once people lost their handsbecause of the traffic accidents or industrial injury accidents, they will get greatinconvenience in their life, they have to face great pain of physical and psychological.Bionic prosthesis arises at the historic moment, the ideal prosthesis design shouldmeet the following two requirements, on the one hand, it must have efficient andreliable control method which can accurately obtain the user’s operation intention, andaccurately control the prosthesis hand to move, on the other hand, it should be able tofeedback the position, the motion state and the surrounding environment informationof prosthesis to human body. Let the users feel indirectly through the prosthesis hand.Based on the topic “The research of motor imagery recognition andproprioceptive feedback on EEG controlled prosthesis”, starting from thebackground and significance of the research, aim at the characteristic of EEG signal,studied and analyzed the acquisition and preprocessing, feature extraction andclassification recognition of motor imagery EEG. Based on proprioceptive feedbackresearch, we present that we can apply kinesthesia illusion induced by vibration asproprioceptive feedback on prosthesis hand to feedback the kinesthesia feeling ofprosthesis to user. This paper mainly completed the following work, and achievedsome innovations:（1）Combined with proprioceptive neuromuscular perception mechanism,introduces the mechanism of kinesthesia illusion induced by high-frequency vibration,and process the experiments on proprioceptive feedback, analysis the characteristicsof the illusion further, especially in determining that when the subjects watchingmoving prosthetic hand, they can still feel the sharp kinesthesia illusion. Discussedthe feasibility of the use of wrist kinesthesia illusion as the feedback of EEGcontrolled prosthetic hand, proposed that kinesthesia illusion can be applied to theprosthetic hand control system that allows users to get the movement sensation ofprosthetic hand in a natural way when they operate the prosthetic hand. Byintroducing wrist motion control mechanism of human, leads EEG prosthetic handand proprioceptive feedback control system design, an analog of human limbmovements closed control loop. And then introduce the composition and experimentalsetup of the system in detail. In addition, in order to explore whether there is a positive impact on motor imagery EEG classification identify by the proprioceptivefeedback, we have designed comparative experiments.（2）For the use of traditional empirical mode decomposition (EMD) method ofEEG de-noising will result in that the useful signal which included in IMF are filteredout, the paper presents a de-noising method based on EMD wavelet threshold method.Such wavelet threshold de-noising is only applied to the high-frequency component ofIMF, rather than a direct effect on the signal, which overcomes the defect thatconventional EMD de-noising cannot retain useful information in the high-frequencycomponent. For the problem that EEG interference by EOG artifact, we propose asecond-order non-stationary sources separation algorithm to eliminate confoundingEOG artifact. The results show that the algorithm is effective removal EOG artifactsfrom EEG.（3）The proposed algorithm using fuzzy entropy to conduct EEG featureextraction, fuzzy entropy algorithm selects the exponential function as a fuzzyfunction to measure the similarity between two vectors, to avoid the approximateentropy and sample entropy function method using binary lack of continuity, thethreshold value for the sensitive, easily lead to the problem of the entropy mutation.Combining the ERD/ERS phenomenon which from motor Imaginary EEG, using C3,C4channel segmentation fuzzy entropy difference as feature vectors, and finallythrough experiments to evaluate the classification accuracy and comparison of results.（4）In EEG pattern classification, at first, introduced linear classificationmethods, and then expounded the basic principle of SVM, finally, discussed theselection problem of penalty factor C and kernel function parameter variable foroptimal support vector machine, focused on the analysis of the genetic algorithmoptimization based SVM classifier works.