Lateral Recognition Of Lower Limbs Based On Somatosensory Evoked Potential And Motor Imagery

Motor imagery(MI)is an imaginary activity of the brain without the physical involvement.The characteristics of electroencephalogram(EEG)generated by MI are mainly distributed in the sensory motor cortex of the scalp,and also have a certain correspondence with the body.The motor imagery can activate certain cerebral cortex,so it has important applications in the field of neurological rehabilitation.The brain-computer interface(BCI)system can convert EEG into control commands.However,In the BCI system composed by MI,the classification of upper limbs is mainly studied.Relatively speaking,due to the overlap of the lower limbs’ projection area in the cerebral cortex,both the feature extraction and classification in lower limbs are difficult.And MI is greatly affected by the individual.Thus there are few studies on the lateral recognition of the lower limbs.In order to realize the classification of lower limbs in MI,the steady-state somatosensory evoked potential(SSSEP)is introduced.It is a steady-state response signal generated in the brain by adding auxiliary stimulation to skin or other susceptors and is mainly concentrated on the fundamental frequency of the stimulation frequency and its harmonics.Combine MI and SSSEP,design rational experimental paradigm,then classify the lower limbs.Firstly,the frequency screening experiment is designed to collect the EEG of the subjects under 5 different stimulation frequency combination,and the spectral characteristics of the signals are analyzed by fast Fourier transform.Then find the optimal stimulation frequency combination that can induce obvious SSSEP.The results show that the subjects have the best spectral characteristics at 28 Hz-33 Hz,and the stimulation intensity for the lower limbs ranged from 10 mA to 25 mA.Design 3 different experimental tasks,they are MI,selective attention,and hybrid paradigm.The 64-channel EEG are collected from the subjects and the results from the frequency screening experiment is used in classical experiment.The analysis of time-frequency maps and brain topographic maps are used to get the characteristics of EEG.The results show that hybrid paradigm with auxiliary stimulation shows more characteristics,comparing the simple motor imagery.The feature extraction is performed by the common spatial mode and the filter band common spatial mode.Then the support vector machine is used for classification,and the accuracy is subjected to ten-fold cross validation.By comparing the averaged classification accuracy of the two feature extraction methods,it is found that the average accuracy of all subjects using common spatial mode in the hybrid paradigm can reach 70.13%,higher than single MI by 4.88%.In the filter band common spatial mode,the average accuracy in the hybrid paradigm is 75.25%,increased by 7.5% than MI.The results show that filter band common spatial mode is more suitable for lateral recognition of the lower limbs.