Influence of different feature selection methods on EMG pattern recognition

Feature extraction is an important method in electromyography (EMG) pattern recognition. High-dimensional EMG features vector lead to redundancy of features. Redundancy of features results in a decrease in classification accuracy of EMG pattern recognition and an increase in computation time for classifier to classify the pattern of EMG signal. Many researchers used feature selection method to decrease the redundancy of features. Sequential forward selection (SFS) and particle swarm optimization (PSO) are widely used in feature selection. This study mainly discusses the effect of two different feature selection methods (SFS and PSO) on EMG pattern recognition. We proposed three methods to compare the different influences of different feature selection methods on EMG pattern recognition. They are support vector machine (SVM) combines with none feature selection method, SVM combines with SFS (SFSSVM) and SVM combines with PSO (PSOSVM). We used SVM, SFSSVM and PSOSVM to classify 11 arm movements respectively. By discussing the classification accuracy and computation time of the three methods, we discussed the different influences of different feature selection methods on EMG pattern recognition. The results showed that the PSOSVM outperformed SVM and SFSSVM. The result implied that PSO is a proper feature selection method for EMG pattern recognition.