Estimation of Grasping Force from Features of Intramuscular EMG Signals with Mirrored Bilateral Training

Abstract

This study investigates the use of features extracted from intramuscular electromyography (EMG) for estimating grasping force in the ipsilateral and contralateral (mirrored) hand, during bilateral grasping tasks. This is relevant since force estimation using mirror tasks is a potentially useful pathway for the clinical training of unilateral amputees. Bilateral grasping force and intramuscular EMG (wire electrodes) of the right forearm were measured in 10 able-bodied subjects. The features extracted from the EMG signal were the root mean square, the global discharge rate, the standard sample entropy, and the constraint sample entropy (CSE). The association between the EMG features and force was modeled using a first-order polynomial model, a second-order exponential model, and an artificial neural network (ANN). The accuracies of estimation of ipsilateral and mirrored grasping force were not significantly different (e.g., R(2) = 0.89 ± 0.02 for ipsilateral and 0.88 ± 0.017 for mirrored, when using CSE and the ANN). It was concluded that it is possible to use just one channel of intramuscular EMG for force estimation. This result suggests that intramuscular EMG signals may be suitable for proportional myoelectric control and that training of the association between intramuscular EMG features and force can be performed using mirror tasks, which is a needed condition for applications in unilateral amputees.