Neural Network-Based Classification of Hand Gestures Using Electromyography Sensor Data

This article focuses on the creation of a machine learning-based method for recognising gestures from electromyography (EMG) signals. In order to categories hand movements such as “thumbs up,” “wave out,” and “hand at rest,” two neural network model were trained and, evaluated using EMG signals obtained from Myo armbands. Traditional back propagation training was used in the first neural network while Bayesian regularisation were use in the second network. The second model performed better with fewer epochs despite having the same number of hidden layers as the first. The findings, indicated that the developed models could correctly categories the various arm movements, with the second model achieving higher level of accuracy. The models’ efficacy in generalising to new, not only untested subjects but also demonstrated by the models’ validation and testing. This study of this paper show how the machine learning and neural network could be used to recognise gestures from EMG signal. These techniques; can be applied to a variety of fields including rehabilitation and human-computer interaction.