A new sEMG signal feature extraction method based on S transform

Abstract

The motion state of the human lower limbs can be reflected by the sEMG signals of the lower limbs muscles. Therefore, extracting features that are distinguishable between different actions can reduce the computational cost of the classification model and improve the accuracy of action classification. This paper proposes a Stockwell transform's singular values concentration measure (ST-SCM) feature extraction method. In order to verify the advantages of the algorithm proposed in this paper, we collected the raw sEMG signals of 10 experimenters' lower limbs movements, and based on these original data, performed verification experiments on the classification performance of ST-SCM. Through three comparative experiments, we compared the classification performance and running time of the ST-SCM feature extraction method with the other five feature extraction algorithms. The experimental results show that the accuracy of this method is 96.19% in the classification of 6 types of single joint movements of the lower limbs, and the running time is relatively fast. Compared with the other five feature extraction algorithms, this feature extraction method can improve the accuracy and reliability of lower limb movement classification.