Research On Online Lower Extremity Motor Imagery Decoding And Sports Rehabilitation Based On FNIRS

Compared with electroencephalogram(EEG),as a subversion of traditional brain-computer interaction,brain-computer interface(BCI)driven by functional near-infrared spectroscopy(fNIRS),is expected to provide an alternative active rehabilitation training method for patients with lower limb dysfunction that seriously affects the quality of life,so it is increasingly favored by researchers.The existing research based on fNIRS-BCI is mostly offline research,and the classification performance needs to be improved.This paper not only improves the classification accuracy based on the fNIRS-BCI system on the basis of offline research,but also builds an online system to decode the lower limb motor imagery task.The paper mainly implements the following tasks:(1)Based on fNIRS and the class-dependent sparse representation classification(cd SRC)decoding walking imagery and idle state research(binary classification).The fNIRS signals of subjects in walking imagery and idle state were collected.After preprocessing the signals,the mean value,peak value,root mean square and their combined features of oxyhemoglobin(Hb O)were extracted.The cd SRC was used to decode the extracted features,and the results of cd SRC were compared with those of support vector machine(SVM),k-nearest neighbor(KNN),linear discriminant analysis(LDA),and logistic regression(LR).The results showed that cd SRC can significantly improve the separability of walking imagery and idle state(p<0.05).This may provide opportunities for rehabilitation training for patients with motor dysfunction.(2)On the basis of the previous binary classification studies,further research is based on fNIRS and cd SRC decoding three classes of gait imagery(normal gait imagery,abnormal gait imagery after stroke,and idle state).The Hb O signals of mean value,peak value,root mean square and their combined features were comprehensively extracted,and the classification performance of different classifiers of cd SRC,SVM and KNN on the three classeses of gait imagery was compared.The results found that the cd SRC decoded of Hb O signals of mean value,peak value and root mean square combined features obtained the highest classification accuracy,which was 87.39 ± 2.59%.Recognizing gait imagery based on fNIRS can be applied to BCI to provide new control commands.This type of BCI can provide the disabled with active rehabilitation training methods,such as providing control commands for mechanical prostheses,so that the disabled can perform active rehabilitation training to restore some of their motor functions.In addition,to our knowledge,this study is the first time that cd SRC has been used to identify three casseses of gait imagery based on fNIRS.(3)On the basis of the above offline study,an online decoding of three classes of lower limb motor imagery based on fNIRS was further studied.The subjects were asked to perform left-leg raising imagery,right-leg raising imagery and idle state.After preprocessing the signals,the Hb O signals of mean value,peak value and root mean square combined features were extracted.The cd SRC was used for classification,and the control commands obtained from the classification were used to control the humanoid robot in real time.The subject’s left-leg imagery,right-leg imagery,and idle state correspond to the humanoid robot moving forward,moving backward,and stopping,respectively.This study is expected to provide assisted rehabilitation training for patients with motor dysfunction,and is also expected to be used to control wheelchairs to provide convenience for patients who have lost motor functions.In addition,this study is the first online decoding of three classes of leg raising motor imagery,and the results show that online decoding of three classes of lower limb motor imagery based on fNIRS-BCI is feasible.