RNN-Based Hand Gesture Recognition Using sEMG

Abstract

Using Deep Learning methods for processing sEMG signals expanded widely in the last decade, giving promising results concerning improving the performance in both directions: evaluation metrics values and fastness of training and testing stages. One of the most common methods in this field are RNN since their structure suits the nature of sEMG signal as a time series signal. In this article, we designed two models using RNN; one was one-layered and the other was multi-layered. We used the original signals values as the features, rather than extracting any other features like RMS, aiming at reducing the complexity of the processing. We used NinaPro-DB5 sEMG dataset. To overcome the obstacle of the dataset being of a small size, which most of such a type of biomedical signals datasets suffer from, we used augmentation by averaging the samples. This approach allowed building classifiers that can classify the samples into the entire 53 classes of NinaPro-DB5 sEMG dataset, with an accuracy of 99.6 and 99.8 % for 1L-RNN and ML-RNN respectively. While the processing time during the test stage ranges were in the range of 60--240 μs per sample for both models, which is fast. These results show that augmentation helped achieve state-of-the-art performance

Please cite this article as:

Assalama L., Potekhin V.V. RNN-based hand gesture recognition using sEMG. Herald of the Bauman Moscow State Technical University, Series Instrument Engineering, 2026, no. 1 (154), pp. 47--58. EDN: FQPDTF

Литература

[1] Cote-Allard U., Fall C.L., Drouin A., et al. Deep learning for electromyographic hand gesture signal classification using transfer learning. IEEE Trans. Neural Syst. Rehabil. Eng., 2019, vol. 27, no. 4, pp. 760--771. DOI: https://doi.org/10.1109/TNSRE.2019.2896269

[2] Chaiyaroj A., Sri-Iesaranusorn P., Buekban C., et al. Deep neural network approach for hand, wrist, grasping and functional movements classification using low-cost sEMG sensors. IEEE BIBM, 2019, pp. 1443--1448. DOI: https://doi.org/10.1109/BIBM47256.2019.8983049

[3] Simao M., Neto P., Gibaru O. EMG-based online classification of gestures with recurrent neural networks. Pattern Recognit. Lett., 2019, vol. 128, pp. 45--51. DOI: https://doi.org/10.1016/J.PATREC.2019.07.021

[4] Koch P., Dreier M., Maass M., et al. RNN with stacked architecture for sEMG based sequence-to-sequence hand gesture recognition. 28th EUSIPCO, 2021, pp. 1600--1604. DOI: https://doi.org/10.23919/Eusipco47968.2020.9287828

[5] Ketyko I., Kovacs F., Varga K.Z. Domain adaptation for sEMG-based gesture recognition with recurrent neural networks. IJCNN, 2019. DOI: https://doi.org/10.1109/IJCNN.2019.8852018

[6] Sosin I., Kudenko D., Shpilman A. Continuous gesture recognition from sEMG sensor data with recurrent neural networks and adversarial domain adaptation. 15th ICARCV, 2018, pp. 1436--1441. DOI: https://doi.org/10.1109/ICARCV.2018.8581206

[7] Hu Y., Wong Y., Wei W., et al. A novel attention-based hybrid CNN-RNN architecture for sEMG-based gesture recognition. PLoS ONE, 2018, vol. 13, no. 10, art. e0206049. DOI: https://doi.org/10.1371/journal.pone.0206049

[8] Rahimian E., Zabihi S., Atashzar S.F., et al. Surface EMG-based hand gesture recognition via hybrid and dilated deep neural network architectures for neurorobotic prostheses. J. Med. Robot. Res., 2020, vol. 5, no. 01n02, art. 2041001. DOI: https://doi.org/10.1142/S2424905X20410019

[9] Samadani A. Gated recurrent neural networks for EMG-based hand gesture classification. A comparative study. 40th EMBC, 2018. DOI: https://doi.org/10.1109/EMBC.2018.8512531

[10] Potekhin V.V., Assalama L. Hand gesture recognition using sEMG with XGBoost, and averaging for augmentation. 26th Int. Sci., Educational and Pract. Conf. System Analysis in Engineering and Control, 2022, vol. 2, pp. 74--88. DOI: https://doi.org/10.18720/SPBPU/2/id23-83

[11] Atzori M., Muller H., Baechler M. Recognition of hand movements in a trans-radial amputated subject by sEMG. IEEE 13th ICORR, 2013. DOI: https://doi.org/10.1109/icorr.2013.6650486

[12] Atzori M., Gijsberts A., Müller H., et al. Classification of hand movements in amputated subjects by sEMG and accelerometers. 36th Ann. Int. Conf. of the EMBC, 2014, pp. 3545--3549. DOI: https://doi.org/10.1109/embc.2014.6944388

[13] Atzori M., Muller H. The Ninapro database: a resource for sEMG naturally controlled robotic hand prosthetics. 37th Ann. Int. Conf. of the IEEE EMBC, pp. 7151--7154. DOI: https://doi.org/10.1109/embc.2015.7320041

[14] Shen S., Gu K., Chen X.R., et al. Gesture recognition through sEMG with wearable device based on deep learning. Mobile Netw. Appli., 2020, vol. 25, no. 6, pp. 2447--2458. DOI: https://doi.org/10.1007/s11036-020-01590-8

[15] Peng F., Chen C., Lv D., et al. Gesture recognition by ensemble extreme learning machine based on surface electromyography signals. Front. Hum. Neurosci., 2022, vol. 16, art. 911204. DOI: https://doi.org/10.3389/fnhum.2022.911204

[16] Li S., Zhang Y., Tang Y., et al. Real-time sEMG pattern recognition of multiple-mode movements for artificial limbs based on CNN-RNN algorithm. Electronics, 2023, vol. 12, no. 11, art. 2444. DOI: https://doi.org/10.3390/electronics12112444

[17] Mora M.C., Garcia-Ortiz J.V., Cerda-Boluda J. sEMG-based robust recognition of grasping postures with a machine learning approach for low-cost hand control. Sensors, 2024, vol. 24, no. 7, art. 2063. DOI: https://doi.org/10.3390/s24072063

[18] Simao M., Neto P., Gibaru O. EMG-based online classification of gestures with recurrent neural networks. Pattern Recognit. Lett., 2019, vol. 128, pp. 45--51. DOI: https://doi.org/10.1016/j.patrec.2019.07.021

[19] Emimal M., Hans W.J., Inbamalar T.M., et al. Classification of EMG signals with CNN features and voting ensemble classifier. Comput. Methods Biomech. Biomed. Eng., 2025, vol. 28, no. 7, 1042--1056. DOI: https://doi.org/10.1080/10255842.2024.2310726

[20] Azhiri R.B., Esmaeili M., Nourani M. Real-time EMG signal classification via recurrent neural networks. IEEE BIBM, 2021, pp. 2628--2635. DOI: https://doi.org/10.1109/bibm52615.2021.9669872