Classification of hand rehabilitation surface electromyographic signals based on FLO-SVM classifier

EMG signals are the electrical activity generated by muscles during contraction and relaxation and contain a wealth of neuromuscular information, and the use of EMG signals for classification is a critical step in the utilization of robotics to achieve rehabilitation goals. Support Vector Machines (SVM) are a powerful machine learning algorithm, but the classification performance of Support Vector Machines is strongly influenced by the choice of its own parameters. In this paper, we propose a pattern recognition method based on the optimization of SVM parameters by the Folded Lizard Algorithm (FLO) and compare the finger movement pattern recognition rate with the models obtained by SVM, Genetic Algorithm (GA) optimized SVM parameters and Sparrow Search Algorithm (SSA) optimized SVM parameters, and validate the proposed method using the NinaPro DB1 dataset. The experimental results show that the recognition rate of the FLO-optimized SVM model is 11.67% higher than that of the unoptimized SVM model, and 2.78% higher than that of the SSA-SVM model, which verifies that the proposed method can improve the accuracy of classification.