Semg signal acquisition: of hand movements for feature extraction and classification

Abstract

This paper explores the classification of surface electromyography (sEMG) signals for hand movement recognition using time and frequency domain features used for feature extraction. Three hand movements were recorded from four healthy subjects for signal analysis achieving an SNR range of 14.81dB to 23.14dB. Two machine learning classifiers, medium neural network (MNN) and cubic support vector machine (CSVM), were evaluated to determine their effectiveness performance. MNN achieved the highest accuracy of 93.8%, demonstrating robust feature separation, while CSVM provided a simpler but slightly less precise result at 92.5%. The findings underscore the potential of MNN in classification of hand movements and aids on the development of advanced prosthetic control systems.