Advanced Machine Learning Proportional Estimation of Muscle Fatigue on Wrist Flexors Using HD-EMG
摘要
This work explores using high-density electromyography (HD-EMG) signals and machine learning models, including Multiple Linear Regression (MLR), Feed Forward Neural Networks (FFNN), and Long Short-Term Memory (LSTM) networks, to quantitatively estimate muscle fatigue as a percentage. MLR and FFNN perform adequately in simpler scenarios but struggle with increased data complexity. In contrast, LSTM shows superior predictive capabilities in correlating HD-EMG signals with muscle fatigue levels. This system suggests great potential for precise and objective fatigue assessment, promising significant advancements in critical fields such as rehabilitation.