基于肌肉协同特征的康复机器人辅助踏步运动关节角度估计

doi:10.3901/JME.2025.15.261

摘要/Abstract

摘要： 针对脑卒中患者肢体与康复机器人交互协同性差和动作不自然的问题，提出一种基于肌肉协同特征(Temporal-spatial muscle synergy feature, TSMS)、双向长短期记忆网络(Bi-directional long short-term memory, BiLSTM)与注意力机制(Attention)四肢多关节角度连续估计算法(TSMS-BiLSTM-Attention)。设计制作床上四肢协同康复机器人，并建立康复机器人上肢和下肢模块的运动学模型。采集人体踏步运动的上肢和下肢表明肌电信号(Surface electromyography,s EMG)和惯性传感器信号(Inertial measuring unit,IMU)。利用非负矩阵分解提取人体踏步运动中的静态肌肉协同特征，并确定4种肌肉协同模式。提出肌肉协同特征的TSMS-BiLSTM-Attention关节角度估计模型，用以估计运动中的四肢关节角度，并采用带异常值的非负矩阵分解算法对s EMG信号实时提取肌肉协同特征。通过IMU测量的欧拉角与四肢运动学模型计算关节角度的测量值。基于TSMS-BiLSTM-Attention模型提取s EMG信号的时空信息，建立四肢s EMG信号与关节角度序列间的映射关系。招募了7名受试者开展踏步运动实验，并整理、分析了相关实验数据。结果表明基于肌肉协同特征的关节角度估计表现显著优于时域、频域和时频特征，关节角度估计的决定系数分别为0.92、0.88、0.86和0.91。对提出的踏步运动关节角度估计模型在床上四肢康复机器人上进行验证，在主动康复训练中，髋关节、膝关节、肩关节、肘关节在线回归的均方根误差分别为3.56°、2.11°、2.36°和3.39°，相对分析误差分别为5.63°、10.13°、7.92°和7.24°。

关键词: 康复机器人, 肢体康复, 表面肌电信号, 肌肉协同, 关节角度估计

Abstract: Aiming at the problems of poor coordination and unnatural movement between limbs of stroke patients and rehabilitation robots, a continuous multi-joint angle estimation method based on temporal-spatial muscle synergy feature（TSMS）, bi-directional long short-term memory（BiLSTM） and attention is proposed. A bed four-limb collaborative rehabilitation robot is designed and developed, and kinematics models of the upper limb and lower limb modules are established. Surface electromyographys（sEMG）and inertial measuring unit（IMU） are collected from the upper and lower limbs of human stepping movement. Non-negative matrix factorization（NMF） is used to extract the static muscle coordination features of human stepping movement, and four muscle coordination patterns are determined. The TSMS-BiLSTM-Attention joint angle estimation model of muscle coordination features is proposed to estimate the joint angle of limbs in the movement, and the NMF algorithm with outliers is used to extract muscle coordination features from sEMG signals in real time. The measured value of joint angle is calculated by using the Euler angle measured by IMU and the kinematic model of limbs. The spatiotemporal information of sEMG signal is extracted based on TSMS-BiLSTM-Attention model, and the mapping relationship between sEMG signal and joint angle sequence is established. Seven subjects are recruited to carry out the step exercise experiment, and the relevant experimental data are sorted out and analyzed. The results show that the joint angle estimation performance based on muscle coordination features is significantly better than the time domain, frequency domain, and time-frequency features, and the determination coefficients of joint angle estimation are 0.92, 0.88, 0.86 and 0.91, respectively. The step movement joint angle estimation model proposed in this paper is verified on the bed four-limb rehabilitation robot. In active rehabilitation training, the root-mean-square errors of hip, knee, shoulder, and elbow joints are 3.56°, 2.11°, 2.36° and 3.39° respectively, and the relative analysis errors are 5.63°, 10.13°, 7.92°, and 7.24° respectively.

Key words: rehabilitation robot, limb rehabilitation, surface electromyography signal, muscle coordination, joint angle estimation

中图分类号:

TP242

冯永飞, 杨圣业, 王琦, 卢衍正, 田俊杰, 王洪波, 牛建业. 基于肌肉协同特征的康复机器人辅助踏步运动关节角度估计[J]. 机械工程学报, 2025, 61(15): 261-274.

FENG Yongfei, YANG Shengye, WANG Qi, LU Yanzheng, TIAN Junjie, WANG Hongbo, NIU Jianye. Joint Angle Estimation of Rehabilitation Robot-assisted Step Movement Based on Muscle Coordination Features[J]. Journal of Mechanical Engineering, 2025, 61(15): 261-274.

参考文献

[1] 赖一楠,叶鑫,丁汉.共融机器人重大研究计划研究进展[J].机械工程学报, 2021, 57(23):1-11. LAI Yinan, YE Xin, DING Han. Research progress of major research plan on Tri-Co robots[J]. Journal of Mechanical Engineering, 2021, 57(23):1-11.
[2] STINEAR J W, BYBLOW W D. Rhythmic bilateral movement training modulates corticomotor excitability and enhances upper limb motricity poststroke:A pilot study[J]. Journal of Clinical Neurophysiology, 2004, 21(2):124-131.
[3] LIANG Xu, YAN Yuchen, WANG Weiqun, et al. Adaptive human-robot interaction torque estimation with high accuracy and strong tracking ability for a lower limb rehabilitation robot[J]. IEEE/ASME Transactions on Mechatronics, 2024, 29(6):4814-4825.
[4] LIANG Xu, HE Guangping, SU Tingting, et al. Finite-time observer based variable impedance control of cable-driven continuum manipulators[J]. IEEE Transactions on Human-Machine Systems, 2022, 52(1):26-40.
[5] 李庆玲,孔民秀,杜志江,等. 5-DOF上肢康复机械臂交互式康复训练控制策略[J].机械工程学报, 2008, 44(9):169-176. LI Qingling, KONG Minxiu, DU Zhijiang, et al. Interactive rehabilitation exercise control strategyfor 5-DOF upper limb rehabilitation arm[J]. Journal of Mechanical Engineering, 2008, 44(9):169-176.
[6] GUATIBONZA A, SOLAQUE L, VELASCO A, et al. Assistive robotics for upper limb physical rehabilitation:A systematic review and future prospects[J]. Chinese Journal of Mechanical Engineering, 2024, 37(1):1-24.
[7] VUJAKLIJA I, Roche A D, HASENOEHRL T, et al. Translating research on myoelectric control into clinics-Are the performance assessment methods adequate?[J]. Frontiers in Neurorobotics, 2017, 11:7.
[8] NING Jiayang, QI An, KOGAMI H, et al. Temporal features of muscle synergies in sit-to-stand motion reflect the motor impairment of post-stroke patients[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2019, 27(10):2118-2127.
[9] D'AVELLA A, SALTIEL P, BIZZI E. Combinations of muscle synergies in the construction of a natural motor behavior[J]. Nature Neuroscience, 2003, 6(3):300-308.
[10] BO Huang, CAI Huaxiong, CHEN Wenbin, et al. Common kinematic synergies of various human locomotor behaviours[J]. Royal Society Open Science, 2021, 8(4):210161.
[11] 黄永欣,文斌,毛菁菁,等.利用肌肉协同研究上肢简单和复杂运动的关联性[J].西安交通大学学报, 2023, 57(9):193-202. HUANG Yongxin, WEN Bin, MAO Jingjing, et al. Using muscle coordination to study the correlation between simple and complex upper limb movements[J]. Journal of Xi'an Jiaotong University, 2023, 57(9):193-2022
[12] SABZEVARI V R, JAFARI A H, BOOSTANI R. Muscle synergy extraction during arm reaching movements at different speeds[J]. Technology and Health Care, 2017, 25(1):123-136.
[13] LI Zhicai, ZHAO Xinyu, WANG Ziyao, et al. A hierarchical classification of gestures under two force levels based on muscle synergy[J]. Biomedical Signal Processing and Control, 2022, 77:103695.
[14] 周文博.基于下肢表面肌电信号的连续运动模态识别研究[D].重庆:重庆大学, 2022. ZHOU Wenbo. Research on continuous motion modal recognition based on surface electromyography signals of lower limbs[D]. Chongqing:Chongqing University, 2022.
[15] 秦涛智.基于sEMG的上肢多关节连续运动估计研究[D].杭州:杭州电子科技大学, 2021. QIN Taozhi. Research on continuous motion estimation of upper limb multi joint based on sEMG[D]. Hangzhou:Hangzhou University of Electronic Science and Technology, 2021.
[16] 陈琪琪,徐乐义,林玲,等.双侧训练对脑卒中患者上肢运动功能及平衡功能的影响分析[J].中国康复, 2022, 37(6):359-362. CHEN Qiqi, XU Leyi, LIN Ling, et al. Analysis of the effects of bilateral training on upper limb motor function and balance function in stroke patients[J]. China Rehabilitation, 2022, 37(6):359-362.
[17] SHI D, ZHANG W, ZHANG W, et al. A review on lower limb rehabilitation exoskeleton robots[J]. Chinese Journal of Mechanical Engineering, 2019, 32:1-11.
[18] VATINNO A A, SCHRANZ C, SIMPSON A N, et al. Predicting upper extremity motor improvement following therapy using EEG-based connectivity in chronic stroke[J]. NeuroRehabilitation, 2022, 50:105-113.
[19] ZHANG Li, LIU Geng, HAN Bing, et al. sEMG based human motion intention recognition[J]. Journal of Robotics, 2019(1):3679174.
[20] SHENG Yixuan, ZENG Jia, LIU Jinbiao, et al. Metric-based muscle synergy consistency for upper limb motor functions[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71:1-11.
[21] 徐翰林,胡国炯,郑绍城,等.脑卒中偏瘫患者膝过伸步态与下肢肌肉激活功能的相关性[J].中国康复理论与实践, 2023, 29(2):151-155. XU Hanlin, HU Guojiong, ZHENG Shaocheng, et al. The correlation between knee hyperextension gait and lower limb muscle activation function in stroke patients with hemiplegia[J]. Chinese Rehabilitation Theory and Practice, 2023, 29(2):151-155.
[22] LIU Yixing, GUTIERREZ-FAREWIK E M. Joint kinematics, kinetics and muscle synergy patterns during transitions between locomotion modes[J]. IEEE Transactions on Biomedical Engineering, 2023, 70(3):1062-1071.
[23] HUG F, VOGEL C, TUCKER K, et al. Individuals have unique muscle activation signatures as revealed during gait and pedaling[J]. Journal of Applied Physiology, 2019, 127(4):1165-1174.
[24] MCDONALD C G, FREGLY B J, O'MALLEY M K. Effect of robotic exoskeleton motion constraints on upper limb muscle synergies:A case study[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2021, 29:2086-2095.
[25] TRESCH M C, SALTIEL P, BIZZI E, The construction of movement by the spinal cord[J]. Nature Neuroscience, 1999, 2(2):162-167.
[26] 王洪安,佘青山,马玉良,等.卒中后上肢肌间协同耦合分析研究[J].传感技术学报, 2020, 33(10):1391-1398. WANG Hongan, SHE Qingshan, MA Yuliang, et al. Research on the collaborative coupling analysis of upper limb muscles after stroke[J]. Journal of Sensing Technology, 2020, 33(10):1391-1398
[27] 赵坤坤.基于肌肉协同的人体运动控制研究及其在脑卒中运动损伤分析中的应用[D].南京:东南大学, 2022. ZHAO Kunkun. Research on human motion control based on muscle coordination and its application in stroke motion injury analysis[D]. Nanjing:Southeast University, 2022.
[28] LAMBERT-SHIRZAD N, VAN DER LOOS H F M. On identifying kinematic and muscle synergies:A comparison of matrix factorization methods using experimental data from the healthy population[J]. Journal of Neurophysiology, 2016, 117(1):290-302.
[29] 常清雅.脑卒中康复运动中肌肉协同特性分析方法研究[D].秦皇岛:燕山大学, 2022. CHANG Qingya. Research on the analysis method of muscle synergy characteristics in stroke rehabilitation exercise[D]. Qinhuangdao:Yanshan University, 2022.
[30] GISBRECHT A, SCHULZ A, HAMMER B. Parametric nonlinear dimensionality reduction using kernel t-SNE[J]. Neurocomputing, 2015, 147:71-82.
[31] SAWERS A, ALLEN J L, TING L H. Long-term training modifies the modular structure and organization of walking balance control[J]. Journal of Neurophysiology, 2015, 114(6):3359-3373.
[32] ZHAO Renbo, TAN V Y F. Online nonnegative matrix factorization with outliers[J]. IEEE Transactions on Signal Processing, 2017, 65(3):555-570.
[33] 刘润程.基于多模态特征融合的谎言行为识别模型的研究与实现[D].北京:北京邮电大学, 2022. LIU Runcheng. Research and implementation of a Lie behavior recognition model based on multimodal feature fusion[D]. Beijing:Beijing University of Posts and Telecommunications, 2022.
[34] GUO Menghao, XU Tianxing, LIU Jiangjiang, et al. Attention mechanisms in computer vision:A survey[J]. Computational Visual Media, 2022, 8(3):331-368.