Integration of the K-Fold Cross-Validation Algorithm for the Classification of Pronator Teres EMG Signals

Abstract

Various studies have shown that people with disabilities require mobility support. Assistive devices such as conventional or electric wheelchairs (joysticks) are commonly used. However, joysticks are only effective if the hands are fully functional. Electromyography (EMG) can be utilised as a navigation signal for people with upper limb amputations. This study aims to identify four left-hand movements using EMG via an Artificial Neural Network (ANN). The main EMG features analysed include Root Mean Square (RMS) and Mean Frequency (MF). Hand movement data is processed, and RMS and MF features are extracted before being fed into the ANN. The results of the analysis show that each movement produces different RMS and MF values; however, RMS and MF alone are insufficient for optimal classification. An ANN with 4 hidden layers and 320 neurons was able to classify hand movements with an accuracy of 77.5%, a precision of 77.9%, and a sensitivity of 77.5%. Improvements in ANN performance can be achieved by adding more data and neurons.