PSD Information Based Fatigue Invariant Hand Gesture Recognition Using sEMG

Surface electromyography (sEMG)-based hand gesture recognition (HGR) is crucial for pattern identification in today's society. Muscle fatigue presents a major challenge in sEMG signal analysis which affects the reliability of sEMG-based systems. This paper proposes a method for fatigue invariant HGR using sEMG signals. An onset-offset detection method is used to remove rest state from the signal, followed by extraction of power spectral density (PSD) information and various features for classification process. The model is trained on non-fatigue data and tested with multiple fatigue levels. It is noticed that accuracy decreases with increasing levels of fatigue for most of the cases. Comparative performance has been performed for various classifiers. The random forest (RF) classifier has achieved the better performance with accuracy (ACC), precision (PRE), recall (REC), and F1-score (F1) of 96.81 %, 96.85%, 96.81 %, and 96.81 %, respectively.