Research On Foot Autonomic Action Recognition Based On Brain Electromyography Fusion

Most stroke patients suffer from motor dysfunction of the foot.Rehabilitation of motor dysfunction is an important research field at present.Currently,motion recognition of electroencephalography(EEG)or electromyography(EMG)is relatively mature in a single mode,but the recognition of different motion modes of the same part has certain disadvantages when EEG is used only.It is easy to be doped with other unit action potential components,and EEG electromyographic feature fusion can improve pattern recognition performance and rehabilitation evaluation by combining EEG with EMG.In this study,by designing the experimental paradigm of foot autonomous movement and subsequent signal feature extraction,the fusion of EEG features and surface electromyographic feature layer was carried out to better realize the classification of different autonomic states of the foot.The main research contents of this paper include:In this study,we designed three kinds of experimental paradigm of foot autonomic movements based on "resting state","foot stretching 15°" and "foot stretching 45°",and simultaneously collected 62-channel autonomic EEG signals and 2-channel surface EMG signals of gastrocnemius and soleus muscles of 15 subjects.The results of feature analysis and classification based on EEG signals of foot autonomic motion were studied.Feature analysis was carried out by combining time-frequency atlas and brain topographic map.Power spectral density(PSD)was used to extract features from EEG signals.Support Vector Machine(SVM)and Random forest(RF)were used to extract features from EEG signals.RF and SVM were used to classify PSD features.The results showed that the average classification accuracy reached 88.1% under RF,which was higher than that under SVM.The classification accuracy of SVM and RF algorithm is 74.6% and 79.6%,respectively.The recognition accuracy of RF method is higher,indicating that RF is more suitable for the classification and recognition of EEG signals of foot autonomic actions.The feature analysis and classification of surface electromyographic signals based on autonomic foot movements were studied.By extracting the time domain features of EMG,SVM and RF were used for binary and triple classification analysis.The binary classification results showed that the highest average classification accuracy reached 90.1% under the SVM and RF algorithms.The average classification accuracy of SVM and RF algorithm is 78.8% and 82.7%,indicating that RF is more suitable for the classification and recognition of electromyographic signals of foot autonomous motion,further confirming the feasibility of SVM and RF algorithms in the study of foot autonomous motion recognition.The autonomic movements of feet were classified and recognized based on the fusion of EEG characteristic signals.EEG features and s EMG features were fused in the feature layer,and SVM and RF algorithms were used for classification recognition.The data showed that both of the two algorithms achieved an average classification accuracy of more than 80%,and RF could achieve a higher classification accuracy.The average classification accuracy of EEG fusion under the two algorithms was higher than that under the corresponding algorithms using EEG or EMG alone.Statistical analysis further indicated that EEG fusion could significantly improve the accuracy of foot autonomous movement recognition,which verified the feasibility of the study on foot autonomous movement recognition based on EEG fusion,and was conducive to the rehabilitation evaluation and treatment of stroke patients.