Classification of EMG-based Hand Gestures using Time and Frequency Domain Features

Classifying hand gestures using electromyography (EMG) data has emerged as a crucial field of study, especially for its uses in human-computer interaction, rehabilitation, and prosthetics. This research presents a method for classifying hand gestures by analyzing Electromyography (EMG) signals using established machine learning algorithms, including Quadratic Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Gaussian Naive Bayes. To identify five distinct gestures—pinch, open, fist, wave-in, and wave-out, three main features were derived from the EMG data namely band power, variance, and FFT power. These features serve as input to the classifiers for gesture recognition. Among the tested algorithms, Quadratic SVM demonstrated superior performance in accurately classifying the gestures compared to KNN and Gaussian Naive Bayes. Additionally, its short training time of 3.17 seconds enhances its potential for real-time applications, making it a strong candidate for systems like prosthetic limb control and robotic manipulation, where fast and reliable gesture detection is essential. This study’s findings suggest that the proposed framework is effective for realtime hand gesture recognition based on EMG signals, offering valuable potential in responsive, control-driven applications.