Upper Limb Motion Recognition Based on Electromyography Signals and Support Vector Machine

Abstract

The increasing requirements of the society to help physically disabilities, the old and the injured individuals to avoid many difficulties to do their dayto-day activities and increase their living condition. The upper limb forearm rehabilitative device has been expected to have a better solution for them. Because the traditional recovery system takes much time to get recovery. In this research, Surface electromyography (sEMG) signals were used as the intention command to movement identification of the upper limb forearm. Six types of major wrist movements collected by placing electrodes on four appointed muscles. Feature extraction was conducted under the Time Domain (TD) Statistical features. Mean Absolute Value (MAV) concluded to be the best feature extraction method for the classification, and it has higher accuracy and the low computational power than other statistical features in order to operate as realtime. The Support Vector Machines (SVM) is designed to conduct classifications task and the model was trained by using a linear classifier. The model operated as a real-time working device and trained the model using 128 features of 128ms. The prediction speed of the model was ~330 obj/sec. The whole process of the model was taken only 131.030ms. The study has obtained zero error rates via the confusion matrix.