EMG-Based Hand Gesture Recognition Using Individual Sensors on Different Muscle Groups

In this study, we introduce a novel classifier for hand gesture recognition based on electromyography (EMG). Our approach utilizes individual EMG sensors placed on various parts of the hand to capture signals related to hand movements. We conducted experiments involving eight healthy subjects, who performed three distinct hand gestures, including complex movements such as flexing, lifting, and grabbing an object. The EMG signals were captured from four channels, and from the acquired data, we extracted eight time-domain features. These features were then used to construct classifiers for the three investigated hand gestures, employing both random forest (RF) and logistic regression (LR) machine learning algorithms. Our results indicate that the RF and the LR classifiers achieved mean accuracies of 0.966 and 0.94, respectively. The high accuracies achieved by our classifiers highlight their reliability and effectiveness in capturing and interpreting hand movements, which open new possibilities for intuitive and precise control systems.