sEMG-based lower-limbs motion recognition using deep learning with one dimensional convolutional neural network

Li Zhang; Hui Li; Qingqi Zhu; Xiaobo Zhang; Hongbin Zhang; Heng Zhang; Chao Lu; Peng Xu; Aibin Zhu; Pingping Wei; Li Zhang; Hui Li; Qingqi Zhu; Xiaobo Zhang; Hongbin Zhang; Heng Zhang; Chao Lu; Peng Xu; Aibin Zhu; Pingping Wei

doi:10.3934/bioeng.2025017

AIMS Bioengineering

2025, Volume 12, Issue 3: 370-382. doi: 10.3934/bioeng.2025017

Previous Article Next Article

Research article

sEMG-based lower-limbs motion recognition using deep learning with one dimensional convolutional neural network

1.
Xi'an Jiaotong University, Xi'an, 710049, China
2.
Honghui Hospital, Xi'an Jiaotong University, Xi'an, 710054, China
3.
Xi'an Key Laboratory of Pathogenesis and Precision Treatment of Arthritis, Xi'an, 710054, China
4.
School of Mechanical Engineering, Northwestern Polytechnical University, Xi'an, 710072, China
5.
School of Mechanical Engineering, Xijing University, Xi'an, 710123, China
6.
Shaanxi Key Laboratory of Intelligent Robots, Institute of Robotics and Intelligent Systems, Xi'an Jiaotong University, Xi'an, 710049, China
7.
State Key Laboratory for Manufacturing Systems Engineering, Xi'an Jiaotong University, Xi'an, 710054, China

Received: 05 June 2025 Revised: 29 June 2025 Accepted: 18 July 2025 Published: 22 July 2025

Surface electromyographic (sEMG)-based motion recognition has been successfully applied in the exoskeleton, human–machine interaction, and rehabilitation engineering. To improve the accuracy of motion recognition and then achieve accurate control of the exoskeleton, this paper is devoted to proposing a one-dimensional (1D) convolutional neural network (1D-CNN) to classify seven daily lower-limb motions from the eight main muscles of human lower limbs. The proposed model architecture mainly consisted of three 1D convolutional layers, three max-pooling layers, and one global average pooling layer. The Teager–Kaiser energy operator was adopted for the starting and ending points detection of sEMG signals. To avoid overfitting, only two features (i.e., mean absolute value (MAV) and root mean square (RMS)) were extracted in this study and used as the input of the proposed model. The results show that the accuracy of motion recognition based on our proposed model had been improved to more than 97.0 ± 0.8% and was higher than that based on the deep neural network (DNN) and support vector machine (SVM) models (95.0 ± 0.6% and 92.8 ± 0.7%). The research results and proposed model of this study are significant for research into exoskeleton control based on sEMG signals.
- surface electromyography,
- one dimensional convolutinal neural network,
- motion recognition,
- lower limb,
- exoskeleton
Citation: Li Zhang, Hui Li, Qingqi Zhu, Xiaobo Zhang, Hongbin Zhang, Heng Zhang, Chao Lu, Peng Xu, Aibin Zhu, Pingping Wei. sEMG-based lower-limbs motion recognition using deep learning with one dimensional convolutional neural network[J]. AIMS Bioengineering, 2025, 12(3): 370-382. doi: 10.3934/bioeng.2025017

Related Papers:

Abstract

Surface electromyographic (sEMG)-based motion recognition has been successfully applied in the exoskeleton, human–machine interaction, and rehabilitation engineering. To improve the accuracy of motion recognition and then achieve accurate control of the exoskeleton, this paper is devoted to proposing a one-dimensional (1D) convolutional neural network (1D-CNN) to classify seven daily lower-limb motions from the eight main muscles of human lower limbs. The proposed model architecture mainly consisted of three 1D convolutional layers, three max-pooling layers, and one global average pooling layer. The Teager–Kaiser energy operator was adopted for the starting and ending points detection of sEMG signals. To avoid overfitting, only two features (i.e., mean absolute value (MAV) and root mean square (RMS)) were extracted in this study and used as the input of the proposed model. The results show that the accuracy of motion recognition based on our proposed model had been improved to more than 97.0 ± 0.8% and was higher than that based on the deep neural network (DNN) and support vector machine (SVM) models (95.0 ± 0.6% and 92.8 ± 0.7%). The research results and proposed model of this study are significant for research into exoskeleton control based on sEMG signals.

Acknowledgments

This research is supported by the Shaanxi Province Postdoctoral Science Foundation (2023BSHGZZHQYXMZZ05).

Conflict of interest

The authors declare no conflicts of interest.

References

[1]	Sun J, Wang Y, Hou J, et al. (2024) Deep learning for electromyographic lower-limb motion signal classification using residual learning. IEEE Trans Neural Syst Rehabil Eng 32: 2078-2086. https://doi.org/10.1109/TNSRE.2024.3403723
[2]	Wang E, Chen X, Li Y, et al. (2024) Lower limb motion intent recognition based on sensor fusion and fuzzy multitask learning. IEEE Trans Fuzzy Syst 32: 2903-2914. https://doi.org/10.1109/TFUZZ.2024.3364382
[3]	Sun B, Cheng G, Dai Q, et al. (2021) Human motion intention recognition based on EMG signal and angle signal. Cogn Comput Syst 3: 37-47. https://doi.org/10.1049/ccs2.12002
[4]	Zhang L, Liu G, Han B, et al. (2020) Assistive devices of human knee joint: a review. Robot Auton syst 125: 103394. https://doi.org/10.1016/j.robot.2019.103394
[5]	Zhang LF, Ma Y, Wang C, et al. (2019) A method for arm motions classification and a lower-limb exoskeleton control based on sEMG singles. IEEE Int Conf Adv Robot Mechatron 2019: 118-123. https://doi.org/10.1109/ICARM.2019.8833708
[6]	Asai K, Takase N (2017) Finger motion estimation based on frequency conversion of EMG signals and image recognition using convolutional neural network. IEEE Int Conf Control Autom Syst 2017: 1366-1371. https://doi.org/10.23919/ICCAS.2017.8204206
[7]	Crepin R, Fall CL, Mascret Q, et al. (2018) Real-time hand motion recognition using sEMG patterns classification. Annu Int Conf IEEE Eng Med Biol Soc 2018: 2655-2658. https://doi.org/10.1109/EMBC.2018.8512820
[8]	Zhao DW, Ye XM, Wang S, et al. (2025) Upper limb human-exoskeleton system motion state classification based on semg: application of CNN-BiLSTM-attention model. Sci Rep 15: 18969. https://doi.org/10.1038/s41598-025-02864-5
[9]	Zhou ZW, Tao Q, Su N, et al. (2024) Lower limb motion recognition based on sEMG and CNN-TL fusion model. Sensors 24: 7087. https://doi.org/10.3390/s24217087
[10]	Shi X, Qin PJ, Zhu JQ, et al. (2020) Feature extraction and classification of lower limb motion based on sEMG signals. IEEE Access 8: 132882-1132892. https://doi.org/10.1109/ACCESS.2020.3008901
[11]	Zhang L, Yan Y, Liu G, et al. (2022) Effect of fatigue on kinematics, kinetics and muscle activities of lower limbs during gait. Proc Inst Mech Eng H 236: 1365-1374. https://doi.org/10.1177/09544119221112516
[12]	Liu G, Zhang L, Han B, et al. (2019) sEMG-based continuous estimation of knee joint angle using deep learning with convolutional neural network. IEEE Int Conf Autom Sci Eng 1: 140-145. https://doi.org/10.1109/COASE.2019.8843168
[13]	Kaiser JF (1990) On a simple algorithm to calculate the ‘energy’ of a signal. Int Conf Acoust Speech Signal Process 1: 381-384. https://doi.org/10.1109/ICASSP.1990.115702
[14]	Li X, Aruin A (2005) Muscle activity onset time detection using teager-kaiser energy operator. Int Conf Eng Med Biol Soc 7: 7549-7552. https://doi.org/10.1109/iembs.2005.1616259
[15]	Gu S, Yue Y, Maple C, et al. (2012) Classification of multi-channels SEMG signals using wavelet and neural networks on assistive robot. Int Conf Ind Inform 2012: 1158-1163. https://doi.org/10.1109/INDIN.2012.6301140
[16]	Chowdhury RH, Reaz MBI, Ali MABM, et al. (2013) Surface electromyography signal processing and classification techniques. Sensors 13: 12431-12466. https://doi.org/10.3390/s130912431
[17]	Chang Y, Wang L, Li M, et al. (2023) Research on gait recognition of surface EMG signal based on MPSO-LSTM algorithm. J Mech Med Biol 23: 2340066. https://doi.org/10.1142/S0219519423400663
[18]	Tu P, Li J, Wang H (2024) Lower limb motion recognition with improved SVM based on surface electromyography. Sensors 24: 3097. https://doi.org/10.3390/s24103097
[19]	Pourmokhtari M, Beigzadeh B (2024) Simple recognition of hand gestures using single-channel EMG signals. Proc Inst Mech Eng H 238: 372-380. https://doi.org/10.1177/09544119231225528
[20]	Shi X, Qin P, Zhu J, et al. (2020) Lower limb motion recognition method based on improved wavelet packet transform and unscented kalman neural network. Math Probl Eng 2020: 5684812. https://doi.org/10.1155/2020/5684812
[21]	Zhang W, Bai Z, Yan P, et al. (2024) Recognition of human lower limb motion and muscle fatigue status using a wearable FES-sEMG system. Sensors 24: 2377. https://doi.org/10.3390/s24072377
[22]	Solnik S, Rider P, Steinweg K, et al. (2010) Teager-Kaiser energy operator signal conditioning improves EMG onset detection. Eur J Appl Physiol 110: 489-498. https://doi.org/10.1007/s00421-010-1521-8
[23]	Hudgins B, Parker P, Scott RN (1993) A new strategy for multifunction myoelectric control. IEEE Trans Biomed Eng 40: 82-94. https://doi.org/10.1109/10.204774
[24]	Englehart K, Hudgins B (2003) A robust, real-time control scheme for multifunction myoelectric control. IEEE Trans Biomed Eng 50: 848-854. https://doi.org/10.1109/TBME.2003.813539
[25]	Ullah A, Ali S, Khan I, et al. (2020) Effect of analysis window and feature selection on classification of hand movements using EMG single. Proc 2020 Intell Syst Conf 3: 400-415. https://doi.org/10.1007/978-3-030-55190-2_30
[26]	Lin M, Chen Q, Yan S (2013) Network in network. CoRR abs/1312.4400: 1-10. https://arxiv.org/pdf/1312.4400
[27]	Kingma DP, Ba J (2014) Adam: A method for stochastic optimization. CoRR abs/1412.6980: 1-15. https://arxiv.org/pdf/1412.6980
[28]	Gautam A, Panwar M, Biswas D, et al. (2020) MyoNet: a transfer-learning-based LRCN for lower limb movement recognition and knee joint angle prediction for remote monitoring of rehabilitation progress from sEMG. IEEE J Transl Eng Health Med 8: 1-10. https://doi.org/10.1109/JTEHM.2020.2972523
[29]	Hao J, Yang P, Chen L, et al. (2019) A gait patterns recognition approach based on surface electromyography and three-axis acceleration signals. IOP Conf Ser Mater Sci Eng 533: 012060. http://doi.org/10.1088/1757-899X/533/1/012060
[30]	Ai Q, Zhang Y, Qi W, et al. (2017) Research on lower limb motion recognition based on fusion of sEMG and accelerometer signals. Symmetry 9: 147. https://doi.org/10.3390/sym9080147

Reader Comments

Your name:*

Email:*
© 2025 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)