Abstract: The ability to track upper extremity movement during activity of daily living has the potential to facilitate the recovery of individuals with neurological or physical injuries. Hence, the use of Surface Electromyography (sEMG) signals to predict upper extremity movement is an area of interest in the research community. A less established technique, Force Myography (FMG), which uses force sensors to detect forearm muscle contraction patterns, is also able to detect some movements of the arm. This paper investigates the comparative performance of sEMG and FMG when predicting wrist, forearm and elbow positions using signals extracted from the forearm only. Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) classifiers were used to evaluate the prediction performance of both FMG and sEMG data. Ten healthy volunteers participated in this study. Under a cross validation across a repetition evaluation scheme, the SVM classifier obtained averaged accuracies of 84.3%, 82.4% and 71.0%, respectively, for predicting elbow, forearm and wrist positions using FMG; while sEMG yielded 75.4%, 83.4% and 92.4% accuracies for predicting the same respective positions. The accuracies obtained using SVM are slightly, but statistical significantly, higher than the ones obtained using LDA. However, the trends on the classification performances between FMG and sEMG are consistent. These results also indicate that the forearm FMG pattern is highly influenced by the change of elbow position, while the forearm sEMG is less subjected to the change. Overall, both forearm FMG and sEMG techniques provide abundant information that can be utilized for tracking the upper extremity movements.

Key words: classification, activity monitoring, upper extremity, limb movement, Surface Electromyography (sEMG), Force Myography (FMG)