Improvement in the classification of EMG signals through a convolutional neural network

Abstract

Abstract The study highlights the crucial role of electromyogram signals (EMG) in recognizing hand and finger movements, and their application in controlling prosthetic limbs. Focusing on the development of human–machine interactions and rehabilitation devices, particularly robotic prostheses. This paper introduces an innovative model utilizing a convolutional neural network (CNN) for classifying fundamental hand grip movements. By converting EMG signals into channel-specific image spectrograms, the model achieved an unprecedented $$100\%$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mn>100</mml:mn> <mml:mo>%</mml:mo> </mml:mrow> </mml:math> accuracy in classifying 6 different movements, evaluated on the EMG hand gesture dataset from the UCI public repository. The results show superior performance compared to advanced methods, demonstrating the model’s potential as a cost-effective and precise control unit for accurately classifying hand grips from EMG signals.