Hindawi Applied Bionics and Biomechanics Volume 2021, Article ID 8850785, 11 pages https://doi.org/10.1155/2021/8850785 Research Article Upper-Limb Muscle Synergy Features in Human-Robot Interaction with Circle-Drawing Movements 1 2 2 3 Cheng Wang , Shutao Zhang , Jingyan Hu , Zhejing Huang , and Changcheng Shi Emergency Trauma Surgical Department, Ningbo First Hospital, Ningbo, Zhejiang 315010, China Cixi Institute of Biomedical Engineering, Ningbo Institute of Materials Technology and Engineering, Chinese Academy of Sciences, Ningbo, Zhejiang 315300, China Rehabilitation Department, Ningbo Yinzhou No. 2 Hospital, Ningbo, Zhejiang 315192, China Correspondence should be addressed to Changcheng Shi; firstname.lastname@example.org Cheng Wang and Shutao Zhang contributed equally to this work. Received 5 September 2020; Accepted 16 August 2021; Published 15 September 2021 Academic Editor: Nan Xiao Copyright © 2021 Cheng Wang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The upper-limb rehabilitation robots can be developed as an eﬃcient tool for motor function assessments. Circle-drawing has been used as a speciﬁc task for robot-based motor function measurement. The upper-limb movement-related kinematic and kinetic parameters measured by motion and force sensors embedded in the rehabilitation robots have been widely studied. However, the muscle synergies characterized by multiple surface electromyographic (sEMG) signals in upper limbs during human-robot interaction (HRI) with circle-drawing movements are rarely investigated. In this research, the robot-assisted and constrained circle-drawing movements for upper limb were used to increase the consistency of muscle synergy features. Both clockwise and counterclockwise circle-drawing tasks were implemented by all healthy subjects using right hands. The sEMG signals were recorded from six muscles in upper limb, and nonnegative matrix factorization (NMF) analysis was utilized to obtain muscle synergy information. Both synergy pattern and activation coeﬃcient were calculated to represent the spatial and temporal features of muscle synergies, respectively. The results obtained from the experimental study conﬁrmed that high structural similarity of muscle synergies was found among the subjects during HRI with circle-drawing movement by healthy subjects, which indicates healthy people may share a common underlying muscle control mechanism during constrained upper-limb circle-drawing movement. This study indicates the muscle synergy analysis during the HRI with constrained circle- drawing movement could be considered as a task for upper-limb motor function assessment. 1. Introduction could be integrated into robotic systems and motor functions (i.e., range of motion, force, velocity, and muscle tone and strength) of upper limbs could be quantitatively detected The upper-limb rehabilitation robots can provide continu- and analyzed [4, 5]. Those assessment results can be obtained ously haptic assistance or resistance to stroke patients with after training and provided to physicians locally or remotely motor impairments in order to help them restore the motor function of upper limbs . Since it was early developed in in order to optimize the rehabilitation therapy. Moreover, some assessment results are also analyzed during the training the 1990s , the upper-limb rehabilitation robot has been process and provided to patients in order to visualize the gradually recognized as an eﬀective medical device which rehabilitation progresses and enhance their training could assist some stroke patients to restore or improve their motivation. motor functions of upper limbs . Besides training, rehabil- itation robots have a potential to be considered as an objec- Circle-drawing is one of typical human-robot interaction (HRI) tasks which are widely applied to quantitatively tive rehabilitation assessment tool due to a plenty of sensors 2 Applied Bionics and Biomechanics 360° 0° Start/End point R = 10 Rocker Clockwise Counterclockwise 180° (a) (b) Figure 1: Setup and experimental design. Participants were asked to keep their upper body stable, and their forearm was tied to the joystick of EULRR. (a) Participants sit on the left side of EULRR to implement circle-drawing movement by using their right hands; (b) the illustration of clockwise and counterclockwise circle-drawing movements with constraint of EULRR system. evaluate upper-limb motor function by using rehabilitation was used as a tool to assist and conﬁne the circle-drawing robots . The ability to accurately implement this move- movements and measure the outcome of HRI tasks. The ment task is related to coordination of both elbow and robot-assisted and constrained circle-drawing tasks for shoulder joints. Therefore, circle-drawing-related kinematic upper-limbs movement were used to increase the consis- tency of muscle synergy features. Both clockwise and coun- (i.e., roundness, area, averaged speed, and jerk) and kinetic (HRI force) parameters measured by motion and force sen- terclockwise circle-drawing tasks were implemented by all sors embedded in the rehabilitation robots have been widely healthy subjects using right hands. The sEMG signals were studied as the potential assessment metrics for upper-limb recorded from six muscles in upper limb, and nonnegative motor functions [4, 6, 7]. However, the muscle synergies matrix factorization (NMF) analysis was utilized to obtain characterized by multiple surface electromyographic muscle synergy information. Both synergy pattern and acti- (sEMG) signals in upper limbs during HRI with circle- vation coeﬃcient were calculated to represent the spatial and drawing movements are rarely investigated. temporal features of muscle synergies, respectively. Recon- Tropea et al. compared the muscle synergies of upper structed sEMG data were compared with the raw data in limbs in stroke patients and healthy subjects and observed order to verify the eﬀectiveness of NMF algorithm. The mus- cle synergy features of upper limb in HRI with two direc- that the diﬀerence can reﬂect the functional deﬁcit induced by the neural damages . Scano et al. clustered the muscle tions of circle-drawing movements were analyzed, and the synergies of stroke patients into ﬁve groups and found a consistency of muscular activation patterns was discussed. deep characterization and relationship with clinical assess- ment methods . The previous studies strongly suggested 2. Methods that muscle synergy analysis may be a potentially promising method for assessing motor function stroke patients. How- 2.1. Participants. Twelve healthy adults (10 males and 2 ever, it remains unclear whether the consistency of normal females and with average ages of 25 ± 1 years old) and two or abnormal muscle synergy patterns in upper limbs for stroke patients (2 females, 67 and 39 years old, Brunnstrom stroke patients is good enough for rehabilitation assess- stages III and IV) were involved in this study, who are all ments. The main challenge is the multiple degree-of- right-hand dominant, with no known neurological diseases, freedom and redundancy for upper-limb movements, which no muscular or skeletal impairments history of the upper may cause a large variation of muscular activation patterns. limbs and the trunks, and no functional abnormalities. Hence, in this study, the end-eﬀector upper-limb reha- Before starting the experimentations, all the subjects signed bilitation robot (EULRR) which was developed in the lab an informed consent. The study was approved by the Applied Bionics and Biomechanics 3 Muscle synergy Synergy Data collection Preprocessing extracting analysis V = W × H Filter Active clockwise m×n m×r r×n Healthy subject Active Remove mean Envelope matrix m×n counterclockwise value Experiment devices Rectify VAF Comparison Circle-drawing with robot-assisted Envelope Synergy number r Synergy Normalized m×r structure Recruitment r×n pattern Extracting sEMG features Extracting muscle synergies Figure 2: The illustration of sEMG data preprocessing and muscle synergy analysis. Ningbo Institute of Materials Technology & Engineering, by holding a handle of joystick with a self-comfortable speed Chinese Academy of Sciences. Informed consent was in diﬀerent directions. Before starting the experiment, sub- acquired from each subject. jects performed a simple learning process under the guid- ance of instructors in order to complete the tasks smoothly. 2.2. EULRR System and sEMG Acquisition Device. The EULRR system mainly consists of motor, belt, reducer, 2.4. sEMG Data Acquisition. During the circle-drawing frame, rocker, sensor, and a tray with a grip in space coordi- tasks, the sEMG signals were recorded from six upper-limb muscles including anterior deltoid (AD), posterior deltoid nates. The system has 5DOF in total: the rocker moves along the three axes; the rotation DOF of tray turns around Z-axis (PD), biceps brachii (BB), triceps brachii (TB), ﬂexor carpi and Y-axis. The movement of X/Y direction is transmitted radialis (FCR), and extensor carpi radialis (ECR). Electrodes to the reducer by the X/Y shaft motor through the belt pul- were placed in accordance with the guidelines of sEMG for ley and then transmitted to the frame by the reducer. The noninvasive assessment of muscles (SENIAM) . Each recorded site was cleaned with alcohol and scrub cream movement of Z direction is transmitted by the Z shaft motor through the pulley to the inside of the screw . A 6-axis before placing the electrodes. All the data were collected at force/torque sensor is attached between the tray and the the sampling rate of 2000 Hz. end of rocker to measure the force/torque exerted by the subjects. sEMG signal acquisition equipment uses TRIGNO 2.5. Muscle Synergy Analysis. The collected sEMG signals were preprocessed according to the following steps before wireless sEMG system (Delsys Inc., Massachusetts, USA) which has 16 4-channel sEMG and acceleration acquisition extracting muscle synergies: band-pass-ﬁltering (20- sensor, wireless transmission range is up to 20 m, sensor 400 Hz), subtracting signal mean values to remove direct current oﬀsets, then rectiﬁed, and enveloped. Each row of delay is less than 500 μs (less than a sampling period), sEMG signal sampling rate is about 2000 Hz, baseline noise is less the preprocessed sEMG matrix (V , where m is the num- m×t ber of muscles and t is the recorded time)  was normal- than 750 nV, it has 16-bit signal resolution, and sensor elec- trode is Ag-AgCl electrode with high conduction eﬃciency. ized with respect to its submaximal  and sampled into 1000 points. Because we rectiﬁed the EMG data, all compo- 2.3. Upper-Limb Circle-Drawing Movement Tasks. After the nents of the synergy are nonnegative is reasonable. NMF sEMG electrode placement, the participants were asked to algorithm [12, 14] was chosen here to extract synergy pat- tern matrix W and activation coeﬃcient matrix H . sit on the left side of the EULRR system and carried out all m×r r×n the tasks by using their right hands in the horizontal plane, So the synergy decomposition as the equation V = m×n as shown in Figure 1. In order to avoid unnecessary muscle W × H . A vector of W represents the relative m×r r×n m×r compensation, they were informed to keep trunk steady and weighting of muscles in each module, and the coeﬃcient only use the upper limbs to complete the full circle-drawing H represents the neural command that speciﬁes how r×n movement by moving the joystick of EULRR, which con- much each synergy will contribute to a total muscular activ- strained the circle radius of 10 cm. All subjects were ity pattern . During the extraction, the number of syn- instructed to carry out a series of trials (10 times per task). ergy vector (r) was increased successively from one to six, Subjects were asked to perform ten counterclockwise and and for each iteration of r, the NMF was repeated 20 times, clockwise circle-drawing movements from the start point and the repetition with the lowest residuals of reconstruction arranged along the circular trajectory in a horizontal plane was selected. 4 Applied Bionics and Biomechanics e mean sEMG signals 10 0.5 0 100 200 300 400 500 600 700 800 900 1000 Samples AD TB PD FCR BB ECR (a) e mean sEMG signals 10 0.5 0 100 200 300 400 500 600 700 800 900 1000 Samples AD TB PD FCR BB ECR (b) Figure 3: The typical sEMG results of six muscles in upper limb during the HRI with circle-drawing movements. (a) The sEMG results for counterclockwise circle-drawing movements; (b) the sEMG results for clockwise movements. human-robot interaction in healthy and stroke subjects. This Various methods have been used to determine the appropriate number of muscle synergies underlying a given study mainly includes sEMG data collection during circle- dataset [16, 17]. The criterion of variance account for drawing movement with the EULRR assisted, sEMG data (VAF) [18–20] was adopted here in the following equation: preprocessing, and muscle synergy extraction and analysis. The ﬁrst two parts can be used to obtain the processed sEMG signals. The muscle synergy features can be obtained ∑ V − V ðÞ i,j r ij VAF = 1 − , ð1Þ by analyzing those processed sEMG data. The procedure of ∑ V i,j ij sEMG data preprocessing and muscle synergy analysis is shown in Figure 2. in which V is the reconstructed EMG matrix and the V is the initial EMG matrix. The number of synergy vectors (N 3. Results ) that suﬃciently recaptured the original EMGs was then deﬁned as the minimum number (r) when VAF exceeded 3.1. sEMG Results during Circle-Drawing Movement. The 90% in more than half of the subjects in both groups. We mean and normalized sEMG signal envelopes of 10 times checked the goodness of reconstruction of global and indi- of counterclockwise and circle-drawing movements are vidual muscle’s EMG at N synergy components, which is shown in Figure 3. The sEMG features in the process of sensitive to both shape and amplitude of the signals . HRI with circle-drawing movements were analyzed. The In summary, the muscle synergy features were analyzed sEMG result for the counterclockwise circle-drawing move- in diﬀerent movement directions of circle-drawing during ments is shown in Figure 3(a). Firstly, the BB and TB were sEMG signals sEMG signals Applied Bionics and Biomechanics 5 e reconstruction AD 731 e reconstruction PD 732 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Samples Samples AD PD e reconstruction BB 733 e reconstruction TB 734 1.2 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Samples Samples TB BB e reconstruction FCR 735 e reconstruction ECR 736 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Samples Samples FCR ECR Vr Figure 4: The comparison of raw and reconstructed sEMG data from factorized matrices. three types of muscle synergies were suﬃcient to reconstruct activated in pairs and showed a negative correlation, which might indicate that TB contraction (relaxation) and BB the original sEMG signal. According to the analysis on two relaxation (contraction) happened simultaneously and BB stroke patients’ data, two muscle synergies can be extracted was activated slightly earlier at temporal domain. Secondly, from the raw sEMG data. the AD and PD were activated in pairs and showed a nega- We used the model of time-invariant synergies to extract the muscle synergies . A typical muscle synergy analysis tive correlation. PD was activated slightly earlier at temporal domain. The sEMG result for the clockwise circle-drawing result of counterclockwise circle-drawing movement tasks movements is shown in Figure 3(b). Firstly, the BB and TB by healthy subjects is shown in Figure 5. The W represents were activated in pairs and showed a negative correlation. synergy patterns, and H represents activation coeﬃcients, TB was activated slightly earlier at temporal domain. Sec- and targeted muscle numbers 1-6 represent AD, PD, BB, TB, FCR, and ECR, respectively. The results demonstrated ondly, the AD and PD were activated in pairs and showed a negative correlation. AD was activated slightly earlier at that the ﬁrst synergy pattern mainly includes AD, and the temporal domain. corresponding activation coeﬃcients were mainly activated at the ending of movements for all subjects. The second syn- ergy pattern mainly includes BB, FCR, and ECR; meanwhile, 3.2. The Results of Muscle Synergy Analysis. In the current the corresponding activation coeﬃcients were mainly acti- study, we used VAF > 90% as the threshold to determine vated at the beginning of movements. The third synergy pat- the number of muscle synergies. The mean VAF of all tern mainly includes PD and TB; meanwhile, the healthy subjects’ 10 times circle-drawing movement demon- strated three muscle synergies can be appropriate for the corresponding activation coeﬃcients were mainly activated at the middle process of movements. sEMG data analysis. In Figure 4, the black solid curves rep- Figure 6 shows a typical muscle synergy analysis result resent the raw sEMG data, and the blue dotted line repre- for clockwise circle-drawing movements conducted by the sents the reconstructed sEMG data. It can be seen that sEMG signals sEMG signals sEMG signals sEMG signals sEMG signals sEMG signals 6 Applied Bionics and Biomechanics e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (a) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (b) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (c) Figure 5: The typical results of muscle synergies of upper limbs for healthy subjects during the HRI with counterclockwise circle-drawing movements. (a) The synergy pattern and activation coeﬃcients for the ﬁrst synergy; (b) the synergy pattern and activation coeﬃcients for the second synergy; (c) the synergy pattern and activation coeﬃcients for the third synergy. The structure of synergy represents the synergy pattern of muscle synergies, and the recruitment mode represents the activation coeﬃcients of muscle synergies. Figure 7 shows a typical muscle synergy analysis result healthy subjects. The results indicated the ﬁrst synergy pat- tern includes AD, and the corresponding activation coeﬃ- for counterclockwise circle-drawing movements imple- cients were mainly activated at the ending of movements. mented by stroke subjects. The ﬁrst muscle synergy pattern The second synergy pattern mainly includes BB, FCR, and includes AD, BB, and FCR which are all ﬂexion muscles, ECR; meanwhile, the corresponding activation coeﬃcients and the corresponding activation coeﬃcients were mainly were mainly activated at the middle process of movements. activated at the ending of movements. The second muscle The third synergy pattern includes PD and TB; meanwhile, synergy pattern includes PD, TB, and ECR which are all the activation coeﬃcients were mainly activated at the begin- extensor muscles, and the corresponding activation coeﬃ- ning of movements. cients were activated from the beginning to the middle W W W m×t m×t m×t H H H r×n r×n r×n Applied Bionics and Biomechanics 7 e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (a) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (b) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (c) Figure 6: The typical results of muscle synergies of upper limbs for healthy subjects during the HRI with clockwise circle-drawing movements. (a) The synergy pattern and activation coeﬃcients for the ﬁrst synergy; (b) the synergy pattern and activation coeﬃcients for the second synergy; (c) the synergy pattern and activation coeﬃcients for the third synergy. The structure of synergy represents the synergy pattern of muscle synergies, and the recruitment mode represents the activation coeﬃcients of muscle synergies. processes of movements. Compared to the results for the responding activation coeﬃcients were activated at the healthy subjects, the curve of activation coeﬃcients for beginning of movements. stroke patients has four peaks which might be induced by impaired muscular function of patients. 4. Discussion Figure 8 shows a typical muscle synergy analysis result for clockwise circle-drawing movements implemented by Upper-limb rehabilitation robot could provide high-inten- stroke subjects. The ﬁrst muscle synergy pattern includes sity, repetitive, task-speciﬁc, and interactive exercises for PD, TB, and ECR which are all extensor muscles, and the stroke patients. The robot could be eﬀective to achieve the corresponding activation coeﬃcients were mainly activated desired training functions, informing the subject to complete at the ending of movements. The second muscle synergy the task as well as enabling them to reduce unnecessary mus- pattern includes BB, TB, FCR, and ECR; meanwhile, the cor- cle activation . Besides the training, the rehabilitation W W m×t m×t m×t H H H r×n r×n r×n 8 Applied Bionics and Biomechanics e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (a) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (b) Figure 7: The typical results of muscle synergies of upper limbs for stroke patients during the HRI with counterclockwise circle-drawing movements. (a) The synergy pattern and activation coeﬃcients for the ﬁrst synergy; (b) the synergy pattern and activation coeﬃcients for the second synergy. The structure of synergy represents the synergy pattern of muscle synergies, and the recruitment mode represents the activation coeﬃcients of muscle synergies. conﬁrmed that high structural similarity of muscle synergies robot also can be developed as an eﬃcient assessment tool for patients’ upper-limb motor functions . was found among the healthy subjects during HRI with The structure of muscle synergy for a speciﬁc task may circle-drawing movement, indicating that the healthy people contain useful information on the residual ability of neuro- may share a common underlying muscle control mechanism muscular control in the poststroke patients. Stroke patients during constrained upper-limb circle-drawing movement. It was found that the muscle activation patterns regarding often have upper-limb problems due to abnormally high spasticity of muscles in the shoulder and elbow joints [25, counterclockwise and clockwise circle-drawing movements 26]. Circle-drawing movement requires the coordination of demonstrated a complementary mode, which indicated that both shoulder and elbow joints with multijoint movements. the activation coeﬃcients of muscle synergies may be This task can be considered as a kind of task-speciﬁc, rhyth- aﬀected by the moving directions. The results of muscle synergy analysis for stroke patients mic, interactive training as well as an objective, reliable means of monitoring the change progress of patient’s demonstrated the number of muscle synergy decreases when upper-limb motor function. Regarding the clinical practice, compared with the healthy subjects. The similar phenome- some kinematic indexes including circle area and roundness non was also found by Cheung et al., and this reduction of can give useful objective information regarding arm function synergy number may be due to the neural function changes of stroke survivors . During robot-assisted rehabilitation after patients’ cortical damage . The activation coeﬃ- process, it needs to promote the patients’ muscle synergy cients of muscle synergy for stroke patients were also found to enhance the biomechanical functions of patients’ upper to be diﬀerent when compared with the healthy subjects. limbs. Muscle synergy is helpful to increase understanding There were more peaks in the activation coeﬃcient curve, of the mechanisms involved in restoration of upper-limb especially during the HRI with counterclockwise circle- function poststroke patients. drawing movements. This feature of activation coeﬃcients In this study, the constrained circle-drawing movements might be related to the abnormal motor function of patients’ for upper limb were used as a task for motor function assess- upper limbs as well as their discontinuity of circle-drawing ment. The results obtained from the experimental studies movements. This study indicates the muscle synergy analysis m×t m×t H H r×n r×n Applied Bionics and Biomechanics 9 e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (a) e structure of synergy 0.5 1 234 56 e recruitment mode 0 100 200 300 400 500 600 700 800 900 1000 Samples (b) Figure 8: The typical results of muscle synergies of upper limbs for stroke patients during the HRI with clockwise circle-drawing movements. (a) The synergy pattern and activation coeﬃcients for the ﬁrst synergy; (b) the synergy pattern and activation coeﬃcients for the second synergy. The structure of synergy represents the synergy pattern of muscle synergies, and the recruitment mode represents the activation coeﬃcients of muscle synergies. during the HRI with constrained circle-drawing movement Thirdly, due to the limited clinical resource, the sam- ple size of stroke patients was only two in this study. The could be considered as a task for upper-limb motor function assessments. trend of decrease of muscle synergy number was found There are still several limitations in this study. Firstly, when the stroke patients were compared with healthy the sEMG signals are normalized to the peak value for a spe- subjects, but there was no statistical evidence due to small ciﬁc task . However, the maximal isometric voluntary sample size. A larger sample size of stroke patients with contraction (MVC) may not represent the real maximal acti- similar disease stage will be considered in the future vating level of muscles during the complex movements . study. The MVC measurements in patients are usually aﬀected by their varying degrees of motor deﬁcits. This might bring a Abbreviations larger intersubject variability . Nevertheless. the sEMG HRI: Human-robot interaction variations between diﬀerent tasks and the same task col- sEMG: Surface electromyography lected at diﬀerent recovery stages in the same patients could NMF: Nonnegative matrix factorization not be intuitively comparable by this normalization method EULRR: End-eﬀector upper-limb rehabilitation robot . In future study, a proper method of sEMG normaliza- AD: Anterior deltoid tion is needed to be considered. PD: Posterior deltoid Secondly, the number of extracted muscle synergies has BB: Biceps brachii been proposed to reﬂect the complexity of motor control TB: Triceps brachii . As mentioned in Methods, we extracted the number FCR: Flexor carpi radialis of muscle synergies by using VAF for all participants. A ECR: Extensor carpi radialis threshold needs to be set by experience. The thresholds SENIAM: Surface EMG for noninvasive assessment of may be diﬀerent between healthy subjects and stroke muscles patients. Therefore, a more objective approach to determine VAF: Variance account for the number of muscle synergy is needed to be further MVC: Maximal voluntary contraction. developed. m×t m×t H H r×n r×n 10 Applied Bionics and Biomechanics  S. T. Zhang, G. K. Zuo, C. C. Shi et al., “The sEMG character- Data Availability istics of human upper limb during circle drawing on EULRR The (Excel) data used to support the ﬁndings of this study system,” in 2017 IEEE International Conference on Cybernetics are available from the corresponding author upon request. and Intelligent Systems (CIS) and IEEE Conference on Robotics, Automation and Mechatronics (RAM), pp. 423–427, Ningbo, Zhejiang, China, 2017. Conflicts of Interest  H. J. Hermens, B. Freriks, C. Disselhorst-Klug, and G. Rau, “Development of recommendations for SEMG sensors and The authors declare that they have no competing interests. sensor placement procedures,” Journal of Electromyography and Kinesiology, vol. 10, no. 5, pp. 361–374, 2000. Authors’ Contributions  D. D. Lee and H. S. Seung, “Learning the parts of objects by non-negative matrix factorization,” Nature, vol. 401, CW and SZ conceptualized the idea, proposed the experi- no. 6755, pp. 788–791, 1999. mental design, and participated in data acquisition and anal-  J. F. Yang and D. Winter, “Electromyographic amplitude nor- ysis and writing of the manuscript at all stages. CS malization methods: improving their sensitivity as diagnostic contributed to development of the conceptualized idea, took tools in gait analysis,” Archives of Physical Medicine and Reha- active part in data acquisition, and revised the manuscript at bilitation, vol. 65, no. 9, pp. 517–521, 1984. all stages. JH and ZH participated in data analysis and  M. C. Tresch, V. C. Cheung, and A. d'Avella, “Matrix factori- revised the manuscript at all stages. All authors read and zation algorithms for the identiﬁcation of muscle synergies: approved the ﬁnal manuscript. Cheng Wang and Shutao evaluation on simulated and experimental data sets,” Journal Zhang contributed equally to this work. of Neurophysiology, vol. 95, no. 4, pp. 2199–2212, 2006.  F. Li, Q. Wang, S. Cao, D. Wu, Q. Wang, and X. Chen, “Lower- References limb muscle synergies in children with cerebral palsy,” in 2013 6th International IEEE/EMBS Conference on Neural Engineer-  C. Duret, A. G. Grosmaire, and H. I. Krebs, “Robot-assisted ing (NER), pp. 1226–1229, San Diego, CA, USA, 2013. therapy in upper extremity hemiparesis: overview of an  D. J. Clark, L. H. Ting, F. E. Zajac, R. R. Neptune, and S. A. evidence-based approach,” Frontiers in Neurology, vol. 10, Kautz, “Merging of healthy motor modules predicts reduced p. 412, 2019. locomotor performance and muscle coordination complexity  M. L. Aisen, H. I. Krebs, and N. Hogan, “The eﬀect of robot- post-stroke,” Journal of Neurophysiology, vol. 103, no. 2, assisted therapy and rehabilitative training on motor recovery pp. 844–857, 2010. following stroke,” Archives of Neurology, vol. 54, no. 4,  G. Torres-Oviedo and L. H. Ting, “Subject-speciﬁc muscle pp. 443–446, 1997. synergies in human balance control are consistent across dif-  C. J. Winstein, J. Stein, R. Arena et al., “Guidelines for adult ferent biomechanical contexts,” Journal of Neurophysiology, stroke rehabilitation and recovery: a guideline for healthcare vol. 103, no. 6, pp. 3084–3098, 2010. professionals from the American Heart Association/American  E. Chiovetto, B. Berret, I. Delis, S. Panzeri, and T. Pozzo, Stroke Association,” Stroke, vol. 47, no. 6, pp. 98–169, 2016. “Investigating reduction of dimensionality during single-joint  N. Nordin, S. Q. Xie, and B. Wünsche, “Assessment of move- elbow movements: a case study on muscle synergies,” Frontiers ment quality in robot- assisted upper limb rehabilitation after in Computational Neuroscience, vol. 7, p. 11, 2013. stroke: a review,” Journal of Neuroengineering and Rehabilita-  V. C. Cheung, A. Turolla, M. Agostini et al., “Muscle syn- tion, vol. 11, no. 1, p. 137, 2014. ergy patterns as physiological markers of motor cortical  H. I. Krebs, M. Krams, D. K. Agraﬁotis et al., “Robotic mea- damage,” Proceedings of the National Academy of Sciences surement of arm movements after stroke establishes bio- of the United States of America, vol. 109, no. 36, markers of motor recovery,” Stroke, vol. 45, no. 1, pp. 200– pp. 14652–14656, 2012. 204, 2014.  V. C. Cheung, A. d'Avella, M. C. Tresch, and E. Bizzi, “Central  T. Krabben, B. I. Molier, A. Houwink, J. S. Rietman, J. H. and sensory contributions to the activation and organization Buurke, and G. B. Prange, “Circle drawing as evaluative move- of muscle synergies during natural motor behaviors,” Journal ment task in stroke rehabilitation: an explorative study,” Jour- of Neuroscience, vol. 25, no. 27, pp. 6419–6434, 2005. nal of Neuroengineering and Rehabilitation, vol. 8, no. 1, p. 15,  G. Torres-Oviedo, J. M. Macpherson, and L. H. Ting, “Muscle synergy organization is robust across a variety of postural per-  A. Schwarz, C. M. Kanzler, O. Lambercy, A. R. Luft, and J. M. turbations,” Journal of Neurophysiology, vol. 96, no. 3, Veerbeek, “Systematic review on kinematic assessments of pp. 1530–1546, 2006. upper limb movements after stroke,” Stroke, vol. 50, no. 3,  A. D’Avella and F. Lacquaniti, “Control of reaching move- pp. 718–727, 2019. ments by muscle synergy combinations,” The Journal of Neu-  P. Tropea, V. Monaco, M. Coscia, F. Posteraro, and S. Micera, roscience, vol. 26, no. 30, pp. 7791–7810, 2016. “Eﬀects of early and intensive neuro-rehabilitative treatment  W. W. Wang, B. C. Tsai, L. C. Hsu et al., “Guidance-control- on muscle synergies in acute post-stroke patients: a pilot based exoskeleton rehabilitation robot for upper limbs: appli- study,” Journal of Neuroengineering and Rehabilitation, vol. 10, no. 1, p. 103, 2013. cation to circle drawing for physiotherapy and training,” Jour- nal of Medical and Biological Engineering, vol. 34, no. 3,  A. Scano, A. Chiavenna, M. Malosio, L. Molinari Tosatti, and pp. 284–292, 2014. F. Molteni, “Muscle synergies-based characterization and clus- tering of poststroke patients in reaching movements,” Fron-  G. Kwakkel, B. J. Kollen, and H. I. Krebs, “Eﬀects of robot- tiers in Bioengineering and Biotechnology, vol. 5, p. 62, 2017. assisted therapy on upper limb recovery after stroke: a Applied Bionics and Biomechanics 11 systematic review,” Neurorehabilitation and Neural Repair, vol. 22, no. 2, pp. 111–121, 2008.  D. G. Kamper, A. N. McKenna-Cole, L. E. Kahn, and D. J. Reinkensmeyer, “Alterations in reaching after stroke and their relation to movement direction and impairment severity,” Archives of Physical Medicine and Rehabilitation, vol. 83, no. 5, pp. 702–707, 2002.  J. P. Dewald, P. S. Pope, J. D. Given, T. S. Buchanan, and W. Z. Rymer, “Abnormal muscle coactivation patterns during iso- metric torque generation at the elbow and shoulder in hemi- paretic subjects,” Brain, vol. 118, no. 2, pp. 495–510, 1995.  L. M. Knutson, G. L. Soderberg, B. T. Ballantyne, and W. R. Clarke, “A study of various normalization procedures for within day electromyographic data,” Journal of Electromyogra- phy and Kinesiology, vol. 4, no. 1, pp. 47–59, 1994.  A. Burden, “How should we normalize electromyograms obtained from healthy participants? What we have learned from over 25 years of research,” Journal of Electromyography and Kinesiology, vol. 20, no. 6, pp. 1023–1035, 2010.  G. T. Allison, R. N. Marshall, and K. P. Singer, “EMG signal amplitude normalization technique in stretch-shortening cycle movements,” Journal of Electromyography and Kinesiology, vol. 3, no. 4, pp. 236–244, 1993.
Applied Bionics and Biomechanics – Hindawi Publishing Corporation
Published: Sep 15, 2021