PartialPowersEMG

This repository is an experiment designed for a graduate course in biomedical control engineering. Following are excerpts from the project report. Please contact me for more information.

Introduction

Previous work [1] in showed that a subject can learn to modulate the powers in specified frequency bands of a surface myographic (sEMG) signal to obtain 2 degree of freedom (DOF) control of a cursor on screen (Figure 1). This is done by integrating the power within a band and converting that value to a vertical or horizontal displace ment of the cursor. In that work, subjects were asked to move the cursor to one of three target circles and did so with a 70-98% hit rate. Similar experiments have shown useful for those who have degraded or no motor control in the lower body because it doesn’t depend on the level of muscular exertion alone, so smaller muscles can be used without fatigue[2]–[5]. However, these frequency bands must be specified offline manually and has only been tested successfully with two DOF. This work aims to increase the number of and optimize the position of the frequency bands to provide more DOF to the user. To test the efficacy of the system, a subject completed a cursor control task with 2 DOF and the autocorrelation of the signal is examined. Then the system will be scaled to arbitrarily many DOF. The sEMG signal in the time domain only captures the gross sum of voltages of a muscle signal in the sensor’s area which is subject to skin resistance and capacitance [6]. This is a tradeoff with the more invasive needle or fine wire EMG which penetrates and directly measures the muscle voltage. For non-clinical human-machine-interfaces, sEMG is more tractable to the user and sensor products won’t be subject to Food and Drug Administration scrutiny because of the large precedent of substantially equivalent devices [7]. For these reasons, sEMG is largely preferable to more invasive procedures for everyday products and motivates this method.

sEMG does not guarantee the measurement of a specific muscle because of electrical crosstalk from muscles in the area [8]. Therefore, we can analyze an sEMG signal with the assumption that itis some combination of signals from all motor units within the sensor’s range (see sec. Future Work for clarification here). By taking the Fourier Transform we decompose the signal into its constituent frequencies which represent the magnitudes of each motor unit. Spikes in the resulting power spectrum would indicate a high amplitude signal in that frequency. Thus, a motor unit, or collection of motor units, that fire at a certain frequency and with larger amplitude than their neighbors would exhibit a spike at that frequency. Although motor units are typically recruited according to the ‘size principle’ to modulate force output, units can be modulated in separate muscles to produce complex movements [9]. For example, Broman et al. show that muscle reflex pathways inhibit individual motor unit ac tivity [10]. The frequency domain of sEMG has yet to be studied in depth, with the focus of research being multi sensor arrays with machine learned classifications [11]–[13]. Adding more sensors increases complexity and cost, so a simpler solution with less sensors is preferred [14]. We hypothesize that placing an sEMG sensor near a surface region with densely innervated motor units would allow us to monitor many muscle signals with one sensor, rather than one sensor for each muscle.

Experiment

For this pilot study a single subject performed all trials. The subject is well-versed in EMG control but has not conducted this experiment in full. Since the long-term purpose of this work is to integrate hand prostheses, we attach a sensor (Delsys Trigno sEMG) to a skin location which would likely bear a prosthetic socket. The flexor digitorum superficialis is an ideal muscle candidate due to its complex control of individual fingers and therefore diverse innervation [15]. The sensor records at 2000hz with a range of 11mV and proprietary filtering to correct motion artefacts and stray electrical noise [16]. The sensor is placed above the flexor digitorum superficialis which is identified by flexing, the site is then marked with ink for future experiments. The subject can monitor the exertion of the muscle to ensure proper placement before the experiment begins.

The test is composed of a max exertion phase and a control phase. In the max exertion phase, the subject is instructed to use a cylindrical grip, and maximally contract their hand around a 3 inch diameter PVC pipe for 3 seconds. During this time, the sEMG readings are fed through the processing pipeline. The maximum values from each band are recorded. This is repeated before every control phase, so fatigue effects are controlled. The control phase consists of blocks of 4 trials: a trial for each combination of low and high target cursors (figure 2). These trials are randomized within each block so that after an arbitrary number of blocks, each trial combination appears an equal amount of times. Here, 10 blocks at a time are run before a max exertion phase is run. The trial is successful if the subject overlaps both cursors with the target cursors at the same time and fails after a 10 second timeout.

References

1. C. Perez-Maldonado, A. S. Wexler, and S. S. Joshi, “Two-dimensional cursor-to-target control from single muscle site sEMG signals,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 18, no. 2, pp. 203–209, Apr. 2010.
2. K. R. Lyons and S. S. Joshi, “Real-time evaluation of a myoelectric control method for high-level upper limb amputees based on homologous leg movements,” Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS, vol. 2016-Octob, pp. 6365–6368, 2016.
3. I. M. Skavhaug, K. R. Lyons, A. Nemchuk, S. D. Muroff, and S. S. Joshi, “Learning to modulate the partial powers of a single sEMG power spectrum through a novel human-computer interface,” Hum. Mov. Sci., vol. 47, pp. 60–69, Jun. 2016.
4. S. Vernon and S. S. Joshi, “Brain-muscle-computer interface: Mobile-phone prototype development and testing,” IEEE Trans. Inf. Technol. Biomed., vol. 15, no. 4, pp. 531–538, Jul. 2011.
5. K. R. Lyons and S. S. Joshi, “Paralyzed subject controls telepresence mobile robot using novel sEMG brain computer interface: Case study,” in IEEE International Conference on Rehabilitation Robotics, 2013, pp. 1–6.
6. S. L. Pullman, “Clinical utility of surface EMG,” Neurology, vol. 55, pp. 171–177, 2000.
7. A. Sastry, “Overview of the US FDA Medical Device Approval Process,” Curr. Cardiol. Rep., vol. 16, no. 6, p. 494, Jun. 2014.
8. C. J. De Luca and R. Merletti, “Surface myoelectric signal cross-talk among muscles of the leg,” Electroencephalogr. Clin. Neurophysiol., vol. 69, no. 6, pp. 568–575, 1988.
9. A. Scano, A. Chiavenna, L. M. Tosatti, H. Müller, and M. Atzori, “Muscle synergy analysis of a hand-grasp dataset: A limited subset of motor modules may underlie a large variety of grasps,” Front. Neurorobot., vol. 12, no. September, Sep. 2018.
10. H. Broman, C. J. De Luca, and B. Mambrito, “Motor unit recruitment and firing rates interaction in the control of human muscles,” Brain Res., vol. 337, no. 2, pp. 311–319, Jul. 1985.
11. E. D. Engeberg, “A physiological basis for control of a prosthetic hand,” Biomed. Signal Process. Control, vol. 8, no. 1, pp. 6–15, Jan. 2013.
12. G. Kanitz, C. Cipriani, and B. B. Edin, “Classification of transient myoelectric signals for the control of multi grasp hand prostheses,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 26, no. 9, pp. 1756–1764, 2018.
13. M. Simao, N. Mendes, O. Gibaru, and P. Neto, “A Review on Electromyography Decoding and Pattern Recognition for Human-Machine Interaction,” IEEE Access, vol. 7, pp. 39564–39582, 2019.
14. Ottobock, “BeBionic Hand,” Ottobock, 2018. [Online]. Available: http://bebionic.com/. [Accessed: 24- Aug-2018].
15. R. Balasubramanian and V. J. Santos, The human hand as an inspiration for robot hand development, vol. 95. Cham: Springer International Publishing, 2014.
16. “TrignoTM Research+ - Delsys.” [Online]. Available: https://www.delsys.com/trigno/research/. [Accessed: 05-Jun-2019].
17. K. R. Lyons and B. W. L. Margolis, “{AxoPy}: A {Python} Library for Implementing Human-Computer Interface Experiments,” J. Open Source Softw., vol. 4, no. 34, p. 1191, 2019.
18. C. J. De Luca, A. Adam, R. Wotiz, L. D. Gilmore, and S. H. Nawab, “Decomposition of Surface EMG Signals,” J. Neurophysiol., vol. 96, no. 3, pp. 1646–1657, 2006.
19. W. Sun, J. Zhu, Y. Jiang, H. Yokoi, and Q. Huang, “One-channel surface electromyography decomposition for muscle force estimation,” Front. Neurorobot., vol. 12, no. MAY, pp. 1–12, 2018.
20. S. L. Kilbreath, R. B. Gorman, J. Raymond, and S. C. Gandevia, “Distribution of the forces produced by motor unit activity in the human flexor digitorum profundus,” J. Physiol., vol. 543, no. 1, pp. 289–296, 2002.