Human Upper Limb Movement Recognition Based on Kernel Principal Component Analysis and Support Vector Machines

Abstract

As an important sensing method of exoskeleton, human intention recognition based on surface electromyography (sEMG) has become a research hotspot in recent years. Many existing researches are based on single gesture or single arm movement, which can not be directly applied to the control of upper limb exoskeleton. Therefore, this paper designs five typical arm carrying movements of upper limbs. Aiming at the problem that classification accuracy and real-time performance are difficult to be compatible, we propose a recognition method combining kernel principal component analysis (KPCA) and support vector machines (SVM). Firstly, the sEMG signal is collected as the information source. After filtering and feature extraction, the redundant information is removed by the KPCA, and then the action classification is realized by SVM. In order to improve the recognition accuracy, we use Cuckoo Search (CS) algorithm to optimize the parameters of KPCA-SVM model. Finally, the experiments are implemented and the results show that the algorithm can achieve 95.4% classification accuracy on the premise of real-time performance, which is an effective method for upper limb movement recognition.