Identification of lower limb muscle fatigue in basketball players based on sEMG signals.

Abstract

Muscle fatigue is an inevitable physiological phenomenon during exercise, which not only leads to a decline in athletic performance but also increases the risk of sports injuries. Therefore, effectively identifying an athlete's muscle fatigue states is of critical importance. This study used the Transformer model to investigate the identification of lower limb muscle fatigue states in basketball players based on surface electromyography (sEMG) signals. The lower limb sEMG signals of 15 basketball players were collected during the experimental process, and the three muscles with higher contribution were selected by combining the muscle synergy analysis method, and then 8 types of feature signals were extracted and fused. The results showed that the Transformer fatigue recognition model based on fused features outperformed the single-feature model in all evaluation metrics. The classification accuracies of the three muscles were 94.28% ± 3.25%, 93.36% ± 3.87% and 94.11% ± 3.28% under the fusion-feature-based condition, respectively. In this paper, LSTM and XGBoost were selected as the comparison models, and the results showed that Transformer significantly outperforms the comparison models in all evaluation metrics, exhibiting stronger robustness and generalization ability.