Event Abstract

Back to Event

Myoelectric control of upper limb prosthesis: current status, challenges and recent advances

NING JIANG1* and Dario Farina1
  • 1 University Medical Center Goettingen, Department of Neurorehabilitation Engineering, Germany

Introduction The surface electromyographic (EMG) signal, which is recorded at skin surface, contains neural control information descending onto muscles. It has been used as the control signals for external devices, such as hand prostheses, for several decades. Over the past 60 years, academic research has progressed to more sophisticated algorithms, mainly focused on pattern recognition-based algorithms. However, none of these academic achievements has been implemented in commercial systems until very recently. We provide an overview of both commercial and academic state-of-the-art (SOA) in this field. We emphasizing that there is a gap between the SOA of the two sectors and a shift of focus in the academia research is imperative. We provide a short discussion of relatively novel academic methods, which include techniques for simultaneous and proportional control (SPC) of multiple degrees of freedom (DOF) and the use of individual motor neuron spike trains for direct control. Industrial SOA The ideal myoelectric control can be achieved when 1) multiple EMG signals are available and can be articulated independently and concurrently; 2) the physiological functions of the muscles generating these EMG signals should match the functions or DOFs to be controlled. In this case, each EMG channel is used and only used to control one ‘natural’ function, and two channels are needed for bidirectional control of one DOF (e.g. open and close hand). In clinical reality, the only scenario that is anywhere close to this ideal condition is with some well-trained TMR patients. For these patients, the reinnervated muscles are surgically made ‘native’ to the functions to be controlled, and they are spatially separated, such that the crosstalk among the channels is minimal [1], [2]. Unfortunately, it is extremely difficult to justify TMR for trans-radial amputees, who are by far the majority of upper extremity amputees. For them, the only practical option is the direct control (DC) approach. To control one DOF, the intensity of muscle activities of two EMG channels, often not the ‘natural one’ w.r.t. the functions, are estimated with simple techniques such as mean absolute value. The locations of the channels need to be carefully selected by prosthetist such that the cross-talk between the two is minimal. When the amputee wishes to switch to another DOF (e.g. from hand to wrist or elbow), a co-activation of both channels is to be registered, usually by a strong co-contraction both muscles. This control scheme is simple and robust enough that it became the only clinically and commercially viable control option for myoelectric prosthesis. However, it is difficult and practically impossible to find more than two signal sites at the residual limb of these amputees. Further, the co-contraction switch is not intuitive, requiring significant amount of mental burden of the user. And the number of DOFs that the commercial devices can operate is limit to a small number (usually < 3). This, among other reasons, makes patient compliance of the current prostheses low [3], [4]. Academia SOA Because of these limitations of the industrial SOA, pattern recognition based (PR) algorithms have been explored in academia for several decades [5]. The key assumption of PR for myoelectric control is that consistent (for same motion) and distinctive (for different motions) signal patterns can be extracted. With this assumption, the continuous feature space of multi-channel EMG can be separated into different regions by a classifier, assigning one region to one motion. When a new data sample becomes available, the classifier would assign it to one of the regions. Contrary to the DC approach, in which there is usually a direct mapping between EMG channels and the articulated functions, in the framework of PR, such functional relationship becomes irrelevant to the user, hidden by the algorithm, i.e. it is a black box approach. The crosstalk among EMG channels, which is the main confounding factor in the DC, becomes a source of information, as long as it is consistent. This black box approach has a benefit over the DC approach such that the control can be more intuitive when the user needs to navigate among several DOFs since no co-contraction-based switching is needed. Further, it relaxes the requirement on the prosthetist to find independent channels, which can be challenging or even impossible for many cases (e.g. small stump area). Decades of research in the PR approach have demonstrated that when proper methods are used, high classification accuracies (>95%) can be reached on a large repertoire of motions (>10 classes) [5], [6]. However, until every recently, none of the pattern recognition systems have been implemented in commercial systems, despite of the well-known limitation of the industrial SOA. At the time of writing this article (end of 2014), only one prosthetic control system used pattern recognition algorithms is provided to the market. The system is shown to be advantageous over the DC approach in patients underwent TMR [7]. However, its benefit for the majority of the upper limb amputees without TMR is not clear, and its clinical and commercial viability remains unclear. There are several reasons for this big gap between the academic and industrial SOA in myoelectric control [8]. First, high offline performance does not automatically lead to good controllability in complex tasks, such as those involving sequentially activating several motions, in which case any miss-classification would likely result in a repeat of the entire sequence. Because the control process is a black box process, it is very difficult for the user to immediately correct or adjust when something goes wrong. Second, while natural movements usually articulate multiple DOFs concurrently, the PR control is sequential, i.e. it is only possible to control one motion at a time, apart from few exceptions of recently reported systems in laboratory settings [9][10]. Even in these latest studies, combined motions have to be part of the training data, which can be impractical when the number of classes is large and the number of potential combinations is even larger. Up to this point, no PR system showed the ability of automatic extrapolation from seen single-function data to un-seen combination data. Last but the least, proportional control is not inherent to pattern recognition algorithms, as it is currently implemented in a post hoc manner by calculating the activity level of all EMG channels after the activated class is determined. Robustness is another major challenge. The following factor were shown to significantly affect the performance level of well-established PR algorithms: varying contraction efforts [11], electrode shift [12], arm posture changes [13], long term variability in EMG [14] [15] and electrode-skin contact impedance [16]. In response to these issues, the literature has seen a shift of focus of research effort, from maximizing the classification accuracy in lab settings to reducing the impact of these factors by adaptive algorithms. Several interesting results have been reported, such as robustness against electrode shift [17], varying effort level [18] and slow feature space change [19]. Recent advances Natural movements are usually simultaneous and proportional control (SPC) of multiple DOFs. To realize this type of control using surface EMG so intuitive control can be achieved, there are recent efforts in utilizing regression-based algorithms. Supervised [20], [21] and un-supervised algorithms [22], [23] have been proposed, analyzed and tested. With this approach, the relationship between the EMG channels and the articulated functions is more explicit than PR algorithms, but not as rigid as in DC approach. This feature, along with the ability of SPC, makes the feedback during online experiment very intuitive and helpful. For example, a small miss-activation of wrist pronation while executing wrist flexion will not affect the ability of the user to execute desired function (flexion), and it is easy to correct by slightly activating supination simultaneously with flexion. This is not possible for PR algorithms, as it is a black box approach and do not allow SPC. As a method just began to appear in the literature very recently, this approach is still in its early stage of development. As promising as it is, its clinical relevance is yet to be established. Another fascinating avenue in myoelectric control is the decoding of motor neuron spike trains directly extracted from the multi-channel surface EMG. Using the neural drive to muscle for prosthesis control implies the decoding of the EMG signal by separating the motor unit action potential shapes from the neural information. Recently, the decomposition of multi-channel surface EMG into the individual spike trains of motor neurons has been proven to be possible [25], and it has been tested and validated on a large set of both simulated [26] and experimental signals [27]. Moreover, this method has been recently implemented online [28], which is a fundamental pre-requisite for any myoelectric control application. The feasibility of this method is also reported with data from TMR patients [29]. The decomposition of the EMG and may provide an extremely highly accurate control approach. Despite promising, the use of motor neuron activity for prosthesis control is still a speculative proposal, with no demonstration of its feasibility in a full online system. Conclusion Muscle signals are still the only commercially and clinically viable biological signals to control upper limb prostheses. Despite several decades of academic research in the field, commercial devices are still controlled almost entirely 60-year old algorithms. This gap between industrial and academic SOA calls for a shift of focus in academia research to provide more intuitive, versatile and robust signal processing algorithm.

Acknowledgements

We acknowledge financial support by the European Commission through FP7 grant MERIDIAN (#280778) and European Research Council Advanced Grant DEMOVE (#267888).

References

[1] O. C. Aszmann, H. Dietl, and M. Frey, “Selective nerve transfers to improve the control of myoelectrical arm prostheses,” Handchirurgie, Mikrochirurgie, Plast. Chir., vol. 40, no. 1, pp. 60–5, Mar. 2008.
[2] P. Zhou, M. M. Lowery, K. B. Englehart, H. Huang, G. Li, L. Hargrove, J. P. A. Dewald, and T. A. Kuiken, “Decoding a new neural machine interface for control of artificial limbs.,” J. Neurophysiol., vol. 98, no. 5, pp. 2974–82, Nov. 2007.
[3] E. A. Biddiss and T. T. Chau, “Upper limb prosthesis use and abandonment: a survey of the last 25 years.,” Prosthet. Orthot. Int., vol. 31, no. 3, pp. 236–57, Sep. 2007.
[4] K. Østlie, I. M. Lesjø, R. J. Franklin, B. Garfelt, O. H. Skjeldal, and P. Magnus, “Prosthesis rejection in acquired major upper-limb amputees: a population-based survey.,” Disabil. Rehabil. Assist. Technol., vol. 7, no. 4, pp. 294–303, Jul. 2012.
[5] E. Scheme and K. Englehart, “Electromyogram pattern recognition for control of powered upper-limb prostheses: State of the art and challenges for clinical use,” J. Rehabil. Res. Dev., vol. 48, no. 6, p. 643, 2011.
[6] K. Englehart, B. Hudgin, and P. A. Parker, “A wavelet-based continuous classification scheme for multifunction myoelectric control,” IEEE Trans. Biomed. Eng., vol. 48, no. 3, pp. 302–311, Mar. 2001.
[7] L. J. Hargrove, B. A. Lock, and A. M. Simon, “Pattern recognition control outperforms conventional myoelectric control in upper limb patients with targeted muscle reinnervation.,” Conf. Proc. ... Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. IEEE Eng. Med. Biol. Soc. Annu. Conf., vol. 2013, pp. 1599–602, Jan. 2013.
[8] N. Jiang, S. Dosen, K.-R. Muller, and D. Farina, “Myoelectric Control of Artificial Limbs—Is There a Need to Change Focus? [In the Spotlight],” IEEE Signal Process. Mag., vol. 29, no. 5, pp. 152–150, Sep. 2012.
[9] A. J. Young, L. H. Smith, E. J. Rouse, and L. J. Hargrove, “Classification of simultaneous movements using surface EMG pattern recognition.,” IEEE Trans. Biomed. Eng., vol. 60, no. 5, pp. 1250–8, May 2013.
[10] A. Ameri, E. N. Kamavuako, E. J. Scheme, K. B. Englehart, and P. A. Parker, “Real-time, simultaneous myoelectric control using visual target-based training paradigm,” Biomed. Signal Process. Control, vol. 13, pp. 8–14, Sep. 2014.
[11] T. Lorrain, N. Jiang, and D. Farina, “Influence of the training set on the accuracy of surface EMG classification in dynamic contractions for the control of multifunction prostheses,” J. Neuroeng. Rehabil., vol. 8, p. 25, 2011.
[12] A. J. Young, L. J. Hargrove, and T. A. Kuiken, “The effects of electrode size and orientation on the sensitivity of myoelectric pattern recognition systems to electrode shift.,” IEEE Trans. Biomed. Eng., vol. 58, no. 9, pp. 2537–44, Sep. 2011.
[13] A. Fougner, E. Scheme, A. D. C. Chan, K. Englehart, and O. Stavdahl, “Resolving the limb position effect in myoelectric pattern recognition.,” IEEE Trans. Neural Rehabil. Syst. Eng., vol. 19, no. 6, pp. 644–51, Dec. 2011.
[14] J. He, D. Zhang, X. Sheng, and X. Zhu, “Effects of Long-Term Myoelectric Signals on Pattern Recognition,” in Intelligent Robotics and Applications, Springer Berlin Heidelberg, 2013, pp. 396–404.
[15] B. Wan, L. Xu, Y. Ren, L. Wang, S. Qiu, X. Liu, X. Liu, H. Qi, D. Ming, and W. Wang, “Study on fatigue feature from forearm SEMG signal based on wavelet analysis,” in 2010 IEEE International Conference on Robotics and Biomimetics, 2010, pp. 1229–1232.
[16] E. A. Clancy, E. L. Morin, and R. Merletti, “Sampling , noise-reduction and amplitude estimation issues in surface electromyography,” J. Electromyogr. Kinesiol., vol. 12, pp. 1–16, 2002.
[17] S. Muceli, N. Jiang, and D. Farina, “Extracting Signals Robust to Electrode Number and Shift for Online Simultaneous and Proportional Myoelectric Control by Factorization Algorithms.,” IEEE Trans. Neural Rehabil. Syst. Eng., vol. 22, no. 3, pp. 623 – 633, Oct. 2014.
[18] J. He, D. Zhang, X. Sheng, S. Li, and X. Zhu, “Invariant Surface EMG Feature Against Varying Contraction Level for Myoelectric Control Based on Muscle Coordination.,” IEEE J. Biomed. Heal. informatics, Jun. 2014.
[19] X. Chen, D. Zhang, and X. Zhu, “Application of a self-enhancing classification method to electromyography pattern recognition for multifunctional prosthesis control.,” J. Neuroeng. Rehabil., vol. 10, no. 1, p. 44, Jan. 2013.
[20] J. L. G. Nielsen, S. Holmgaard, N. Jiang, K. B. Englehart, D. Farina, and P. A. Parker, “Simultaneous and proportional force estimation for multifunction myoelectric prostheses using mirrored bilateral training,” IEEE Trans. Biomed. Eng., vol. 58, no. 3, pp. 681–688, Mar. 2011.
[21] J. M. Hahne, F. Biessmann, N. Jiang, H. Rehbaum, D. Farina, F. C. Meinecke, K.-R. Muller, and L. C. Parra, “Linear and nonlinear regression techniques for simultaneous and proportional myoelectric control.,” IEEE Trans. neural Syst. Rehabil. Eng., vol. 22, no. 2, pp. 269–79, Mar. 2014.
[22] N. Jiang, K. B. Englehart, and P. A. Parker, “Extracting simultaneous and proportional neural control information for multiple-DOF prostheses from the surface electromyographic signal,” IEEE Trans. Biomed. Eng., vol. 56, no. 4, pp. 1070–1080, Apr. 2009.
[23] N. Jiang, H. Rehbaum, I. Vujaklija, B. Graimann, and D. Farina, “Intuitive, Online, Simultaneous and Proportional Myoelectric Control Over Two Degrees of Freedom in Upper Limb Amputees.,” IEEE Trans. neural Syst. Rehabil. Eng., vol. 22, no. 3, pp. 501–510, Aug. 2014.
[24] A. Holobar and D. Zazula, “Multichannel Blind Source Separation Using Convolution Kernel Compensation,” IEEE Trans. Signal Process., vol. 55, no. 9, pp. 4487–4496, Sep. 2007.
[25] A. Holobar, D. Farina, M. Gazzoni, R. Merletti, and D. Zazula, “Estimating motor unit discharge patterns from high-density surface electromyogram.,” Clin. Neurophysiol., vol. 120, no. 3, pp. 551–62, Mar. 2009.
[26] A. Holobar, M. A. Minetto, A. Botter, F. Negro, and D. Farina, “Experimental analysis of accuracy in the identification of motor unit spike trains from high-density surface EMG.,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 18, no. 3, pp. 221–9, Jun. 2010.
[27] V. Glaser, A. Holobar, and D. Zazula, “Real-Time Motor Unit Identification From High-Density Surface EMG.,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 21, no. 6, pp. 949 – 958, 2013.
[28] D. Farina, H. Rehbaum, A. Holobar, I. Vujaklija, N. Jiang, C. Hofer, S. Salminger, H.-W. van Vliet, and O. C. Aszmann, “Noninvasive, accurate assessment of the behavior of representative populations of motor units in targeted reinnervated muscles.,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 22, no. 4, pp. 810–9, Jul. 2014.

Keywords: myoelectric control, pattern recognition, Simultaneous and proportional control, motor unit, Upper Limb Prosthesis

Conference: MERIDIAN 30M Workshop, Brixen, Italy, 25 Sep - 25 Sep, 2014.

Presentation Type: Oral Presentation

Topic: Neuroengineering

Citation: JIANG N and Farina D (2014). Myoelectric control of upper limb prosthesis: current status, challenges and recent advances. Front. Neuroeng. Conference Abstract: MERIDIAN 30M Workshop. doi: 10.3389/conf.fneng.2014.11.00004

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 06 Nov 2014; Published Online: 06 Nov 2014.

* Correspondence: Dr. NING JIANG, University Medical Center Goettingen, Department of Neurorehabilitation Engineering, Goettingen, 37075, Germany, ning.jiang@ieee.org