Electromyography(EMG)can reflect the movement intention of the human and it is one of the main signals for exoskeleton and prosthetic control.However,inter-subject variability increases the cost of using surface electromyo-graphy(sEMG)-based discrete hand motion recognition.In response to this situation,a transfer learning method is proposed based on small adjustment sets from the perspective of domain adaptation.This method utilizes Riemann Procrustes analysis(RPA)to extract Riemannian features and traditional time-domain features as the input features of support vector machine(SVM),and its recognition accuracy is verified by experiments.Experiments are carried out on ten subjects,and the Riemann-Plucker analysis under the Riemann feature increases the accuracy by 5％to 7％,compared with the action recognition method without transfer learning.In terms of feature space distribution,the overlap of Riemannian features after Riemann-Plucker analysis is higher.The results show that this method has obvious advantages in recognition of discrete hand movements based on EMG signals.
维普
万方数据