To address the individual differences in fatigue tolerance and the complexity of fatigue characteristics among different muscles in surface electromyography(sEMG)signal-based fatigue detection

a multi-scale spatiotemporal fusion network architecture is proposed to comprehensively characterize fatigue features in sEMG signals by integrating multiple spatiotemporal feature modeling modules.Furthermore

a spatiotemporal neural architecture search method is introduced to automatically construct multi-scale spatiotemporal fusion network for muscle fatigue detection within a heterogeneous module search space.Experimental results on lower-limb muscle fatigue detection demonstrate that the multi-scale spatiotemporal fusion network searched by spatiotemporal neural architecture search method achieves an accuracy of 99.69%

outperforming various comparative methods and providing a reliable technical reference for clinical auxiliary diagnosis.