Mechanomyography (MMG) enables non-invasive monitoring of muscle mechanical activity, while its utility in time-resolved fatigue detection during dynamic human–exoskeleton interaction remains underexplored. This pilot study explored the feasibility of combining MMG with machine learning to characterize neuromuscular fatigue and evaluate passive lower-limb exoskeleton assistance during repetitive 10 kg squat-lifting tasks. MMG signals from five lower-limb muscles were extracted for time-domain, frequency-domain and nonlinear features, and fatigue identification was implemented via a spectral-based criterion and multi-muscle voting optimization. A radial basis function-enhanced random forest (RBF-RF) model integrated with data augmentation was validated through leave-one-subject-out cross-validation. The results demonstrated that the 1/5 voting rule achieved optimal performance, with mean accuracy of 0.913 ± 0.057, AUC of 0.792 ± 0.073, and a low fatigue detection error of 1.4 ± 0.8 s. Data augmentation steadily improved model robustness, and predicted fatigue levels were significantly correlated with subjective perceived exertion (ρ = 0.756, p < 0.001). This pilot study demonstrates the feasibility of MMG-based fatigue monitoring for wearable assistive systems. The proposed framework supports objective, high-temporal-resolution fatigue monitoring, and may serve as a viable tool for assessing wearable assistive systems. Further large-cohort studies are required to validate its generalizability for practical applications.
A novel hand gesture recognition framework based on surface electromyography (sEMG) is proposed for soldier operational scenarios under small-sample conditions. The framework integrates Empirical Mode Decomposition (EMD) for signal reconstruction, histogram-based features, and the Deep Forest (DF) classifier. Evaluations are conducted under two protocols: subject-wise evaluation and mixed-subject nested 8-fold cross-validation. Under subject-wise evaluation, the proposed EMD-HIST-DF method achieves 99.94% accuracy with 0.00027 ms per sample. Under mixed-subject nested 8-fold cross-validation, 98.41% accuracy is maintained with 0.00053 ms per sample. Ablation studies confirm the significant contribution of EMD-based signal enhancement in the mixed-subject setting (approximately 10.6 percentage points, p < 0.001). Parameter sensitivity analysis guides optimal parameter selection, and statistical tests confirm significant performance gains over baseline methods. Confusion matrices illustrate high per-class accuracy with minimal inter-class confusion. The framework shows potential as a promising solution for accurate, efficient, and sample-sparing gesture recognition in resource-constrained environments such as supernumerary robotic limb control.