Effects of Forearm Muscle Fatigue on Classification Performance in sEMG-Based Hand Gesture Recognition

Abstract

Hand gestures can be identified by using surface electromyography (sEMG) of the upper limb, which measures the electrical activity of skeletal muscles. Hand gesture recognition has recently become most commonly used methods in many applications, such as brain computer interfaces, orthosis and prosthesis control etc. However, there are many factors affecting this recognition in the experimental staps and muscle fatigue is one of these factors. In this study, the effect of forearm muscle fatigue on the performance of hand gesture recognition is investigated using sEMG signals. Four healthy subjects perform six hand motions (fist, hold cup, pointing, pinch, open hand, and rest position). Signals were collected under two conditions: with and without muscle fatigue, and the classification performance of each condition was compared. The four statistical features, including mean frequency, wavelength, mean absolute value, and Willison amplitude are extracted from the sEMG signals. Three classification algorithms have been used, which are the Random Forest, the Support Vector Machine, and the Artificial Neural Network. The results show that in sEMG-based classification applications, data collection steps should be performed by considering the fatigue levels of the muscles.