Research On The Technology Of Motor Imagery Regulated Steady-State Somatosensory Evoked Potential Based Brain-Computer Interface

Motor Imagery based Brain-computer Interface(MI-BCI)is an information communication way that does not depend on peripheral nerve and muscle tissues.BCI based on brain motor-related mental activity decoding which converts subjective motor intentions into output instructions,can not only establish an organic information link between motor intentions and limb movements but also be conducive to the rehabilitation of patients with motor disorders.However,the related BCI technology still faces a challenge,that is,low decoding efficiency.In recent years,a new BCI paradigm,Motor Imagery Regulated Steady-State Somatosensory Evoked Potential(MI-SSSEP)has been introduced into the field of MI decoding,which can effectively improve the MI information carried in EEG signals and obtain better decoding results.According to the characteristics of this paradigm,this thesis further develops the relevant feature extraction and recognition algorithms to give full play to the new MISSSEP paradigm and obtain better MI decoding results.Firstly,a new EEG index,Inter-Stimulus Phase Coherence(ISPC),was built to measure phase synchronization in the steady-state evoked potentials in this thesis.12 subjects participated in our experiment.ISPC analysis found that left-handed MI can cause a significant decrease in phase synchronization in contralateral sensorimotor SSSEP,while right-handed MI had little effect on it,and vice versa.Combining ISPC features with traditional spectral power features,the single-channel left-hand versus right-hand MI recognition accuracy reached 81.25%,much higher than the accuracy of traditional MI paradigms which was 61.5%.Secondly,to optimize the high-dimensional features of EEG signals in motor imagery decoding.A new method named Riemann Kernel Support Vector Machine Recursive Feature Elimination(RKSVM-RFE)was proposed based on the manifold information on electroencephalogram(EEG).The EEG data of 10 subjects were collected when they were imagining 7-class movements of different parts of the body.The data was modeled using RKSVM-RFE to recognize the motor intention corresponding to the EEG data.Results show that accuracy from our method is about7% higher than the state-of-the-art method named CSP.And RKSVM-RFE can reduce the complexity of the system because it can decrease 50% EEG channels.Finally,a multi-instruction MI-BCI based on MI-SSSEP and RKSVM-RFE was designed.It can decode eight kinds of MI tasks including motor imagery of compound limb and different force load levels.The average recognition rate of 8 subjects reached77%,and the highest was 89%.The research shows that MI-SSSEP paradigm has rich neurophysiological connotations and great potential in decoding motor intentions.The development of new features and suitable algorithms plays a key role in giving full play to the advantages of MI-SSSEP paradigm,which deserves further study.