Real-time muscle fatigue monitoring via surface electromyography (sEMG) is essential for adaptive human-robot collaboration, yet deploying deep learning models on resource constrained platforms remains challenging. Existing methods typically process all available muscle channels using heavyweight architectures, incurring unnecessary computational overhead and potentially introducing noise. In this paper, we propose a lightweight sEMG fatigue classification framework that combines mutual information-based channel ranking with an Efficient Channel Attention-augmented Residual Network (ECA-ResNet). The data-driven channel ranking module identifies the most discriminative subset from high-density electrode arrays without requiring task-specific anatomical priors, reducing input dimensionality. The backbone network stacks compact 1D residual blocks, each equipped with an ECA module that models crosschannel dependencies via adaptive kernel-size 1D convolution, achieving effective channel recalibration with negligible parameter increase. Experiments on a dynamic lower-limb fatigue dataset (8 subjects, 12 muscles, 3 fatigue levels) demonstrate that the proposed method achieves 79.62% subject-wise accuracy, outperforming attention-based and attention-free deep learning baselines while requiring only 211K trainable parameters and 33% fewer sensors. These results confirm that selective channel utilization combined with efficient attention yields a practical, high-performance solution for embedded sEMG fatigue perception.

Sipei Liu. Lightweight ECA-ResNet with Automated Channel Selection for Real-Time sEMG Fatigue Classification. International Journal of Neural Network (2026), Vol. 5, Issue 1: 22-32. https://doi.org/10.38007/NN.2026.050103

[1] M. R. Al-Mullah, W. S. Selia, and H. P. Lee, “A review of non-invasive techniques to detect and predict localised muscle fatigue,” Sensors, vol. 11, no. 4, pp. 3545–3594, 2011.

[2] S. Hasan and N. Alam, “Comprehensive Comparative Analysis of Lower Limb Exoskeleton Research: Control, Design, and Application,” Actuators, vol. 14, no. 7, p. 342, 2025.

[3] M. Lorenzini, M. Lagomarsino, L. Fortini, S. Gholami, and A. Ajoudani, “Ergonomic human-robot collaboration in industry: A review,” Frontiers in Robotics and AI, vol. 9, p. 813907, 2023.

[4] M. Gonzalez-Izal, A. Malanda, E. Navarro-Amezqueta, and M. Izquierdo, “EMG spectral indices and muscle power fatigue during dynamic contractions,” Journal of Electromyography and Kinesiology, vol. 22, no. 2, pp. 233–240, 2012.

[5] M. Cifrek, V. Medved, S. Tonkovi , and S. Ostoji , “Surface EMG based muscle fatigue evaluation in biomechanics,” Clinical Biomechanics, vol. 24, no. 4, pp. 327–340, 2009.

[6] M. Shariatzadeh, E. Hadizadeh Hafshejani, C. J. Mitchell, M. Chiao, and D. Grecov, “Predicting muscle fatigue during dynamic contractions using wavelet analysis of surface electromyography signal,” Biocybernetics and Biomedical Engineering, vol. 43, no. 2, pp. 428–441, 2023.

[7] A. Moniri, D. Terracina, J. Rodriguez-Manzano, and P. Georgiou, “RealTime Forecasting of sEMG Features for Trunk Muscle Fatigue Using Machine Learning,” IEEE Transactions on Biomedical Engineering, vol. 68, no. 2, pp. 718–727, 2021.

[8] J. Sun, G. Liu, Y. Sun, K. Lin, Z. Zhou, and J. Cai, “Application of surface electromyography in exercise fatigue: A review,” Front. Syst. Neurosci., vol. 16, p. 893275, 2022.

[9] M. Atzori, M. Cognolato, and H. Mu… ller, “Deep learning with convolutional neural networks applied to electromyography data: A resource for the classification of movements for prosthetic hands,” Frontiers in Neurorobotics, vol. 10, p. 9, 2016.

[10] J. Wang, Y. Sun, and S. Sun, “Recognition of Muscle Fatigue Status Based on Improved Wavelet Threshold and CNN-SVM,” IEEE Access, vol. 8, pp. 207914–207922, 2020.

[11] J. Liu, Q. Tao, and B. Wu, “Dynamic Muscle Fatigue State Recognition Based on Deep Learning Fusion Model,” IEEE Access, vol. 11, pp. 95079–95091, 2023.

[12] D. Mu, J. Wang, F. Li, W. Hu, and R. Chen, “Multilevel attention mechanism for motion fatigue recognition based on sEMG and ACC signal fusion,” PLOS ONE, vol. 19, no. 11, p. e0310035, 2024.

[13] G. Hajian and E. Morin, “Deep multi-scale fusion of convolutional neural networks for EMG-based movement estimation,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 30, pp. 486–495, 2022.

[14] S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, “CBAM: Convolutional block attention module,” in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 3–19.

[15] J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 7132–7141.

[16] Q. Wang, B. Wu, P. Zhu, P. Li, W. Zuo, and Q. Hu, “ECA-Net: Efficient channel attention for deep convolutional neural networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 11534–11542.

[17] W. Wu and X. Gu, “A novel attention-guided ECA-CNN architecture for sEMG-based gait classification,” Math. Biosci. Eng., vol. 20, no. 4, pp. 7140–7153, 2023.

[18] A. Phinyomark, P. Phukpattaranont, and C. Limsakul, “Feature reduction and selection for EMG signal classification,” Expert systems with applications, vol. 39, no. 8, pp. 7420–7431, 2012.

[19] A. Kraskov, H. Sto… gbauer, and P. Grassberger, “Estimating mutual information,” Physical Review E, vol. 69, no. 6, p. 066138, 2004.

[20] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778.

[21] A. K. Jain, R. P. W. Duin, and J. Mao, “Statistical pattern recognition: A review,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 1, pp. 4–37, 2000.

[22] C. Cortes and V. Vapnik, “Support-vector networks,” Machine Learning, vol. 20, no. 3, pp. 273–297, 1995.

[23] L. Breiman, “Random Forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001.