SEMG Based Wrist Movement Recognition with Portable Sensing Device

Surface Electromyography (sEMG) based movement recognition on have been applied in many areas. However, sEMG signals are weak signals and can be easily polluted by various environmental noise during the acquisition process, which induces a limited classification accuracy. In order to improve the classification accuracy and enhance the stability of the classifiers, LDA has been applied by many researches. However, the classification performance varies due to different signal preprocessing and feature extraction methods. In this study, we combined LDA with template matching (TM) to solve multi-category classification task for intact subjects. The experimental results show that the classification accuracy of the proposed algorithm reaches 97.5% for 8 wrist motions, and it is better than the classification result of template matching classifier with the same data set. The recognition accuracy of LDA with TM is similar to that of the two algorithms SVM and Adaboost-SVM, but LDATM can save a significant amount of training time cost and testing time cost.