Skip to content
Licensed Unlicensed Requires Authentication Published by De Gruyter January 11, 2017

Convolutive blind source separation of surface EMG measurements of the respiratory muscles

  • Eike Petersen EMAIL logo , Herbert Buchner , Marcus Eger and Philipp Rostalski


Electromyography (EMG) has long been used for the assessment of muscle function and activity and has recently been applied to the control of medical ventilation. For this application, the EMG signal is usually recorded invasively by means of electrodes on a nasogastric tube which is placed inside the esophagus in order to minimize noise and crosstalk from other muscles. Replacing these invasive measurements with an EMG signal obtained non-invasively on the body surface is difficult and requires techniques for signal separation in order to reconstruct the contributions of the individual respiratory muscles. In the case of muscles with small cross-sectional areas, or with muscles at large distances from the recording site, solutions to this problem have been proposed previously. The respiratory muscles, however, are large and distributed widely over the upper body volume. In this article, we describe an algorithm for convolutive blind source separation (BSS) that performs well even for large, distributed muscles such as the respiratory muscles, while using only a small number of electrodes. The algorithm is derived as a special case of the TRINICON general framework for BSS. To provide evidence that it shows potential for separating inspiratory, expiratory, and cardiac activities in practical applications, a joint numerical simulation of EMG and ECG activities was performed, and separation success was evaluated in a variety of noise settings. The results are promising.


We thank Philippe Jolliet and Lise Piquilloud at CHUV Lausanne for providing the clinical data sample used for the evaluation of the proposed algorithm.

Appendix A: Details of the EMG simulation model

The simulated diaphragm consisted of 300 MUs, with 20% being fast-twitch, fatigable MUs and 80% slow-twitch, fatigue resistant MUs. Both of the transversus abdominis muscles consisted of 100 MUs, with 60% being fast-twitch, fatigable MUs and 40% slow-twitch, fatigue resistant MUs. Fast-twitch MUs consisted of between 80 and 160 muscle fibers, while slow-twitch MUs consisted of between 40 and 80 muscle fibers (and had a lower conduction velocity and action potential amplitude).

MU locations were distributed uniformly in a prescribed muscle area. Each MU was then assigned a circular MU territory based on the number and type of muscle fibers belonging to the MU. This MU distribution strategy results in overlapping MU areas, which is a desirable property of models for MU geometry [15].

For MU rate coding and recruitment, a three-compartment piecewise linear model was employed that respects all of the empirical findings presented by De Luca and Erim [7]. Firing rates were assumed to lie in the range between 5 imp/s and 50 imp/s for the diaphragm, and between 5 imp/s and 70 imp/s for the transversus muscles. The mathematical details of the rate coding model may be the subject of a future publication; for the present investigation, however, the details of the rate coding model are irrelevant.

The EMG simulation was implemented in the R programming language [26].


[1] Bartolo A, Roberts C, Dzwonczyk RR, Goldman E. Analysis of diaphragm EMG signals: comparison of gating vs. subtraction for removal of ECG contamination. J Appl Physiol 1996; 80: 1898–1902.10.1152/jappl.1996.80.6.1898Search in Google Scholar

[2] Buchner H, Aichner R, Kellermann W. Blind source separation for convolutive mixtures: A unified treatment. In: Huang Y, Benesty J, editors. Audio signal processing for next-generation multimedia communication systems. Boston: Kluwer 2003.Search in Google Scholar

[3] Buchner H, Aichner R, Kellermann W. Trinicon-based blind system identification with application to multiple-source localization and separation. In: Makino S, Lee T-W, Sawada S, editors. Blind speech separation. Springer: Berlin, Germany 2007.10.1007/978-1-4020-6479-1_4Search in Google Scholar

[4] Buchner H, Petersen E, Eger M, Rostalski P. Convolutive blind source separation on surface EMG signals for respiratory diagnostics and medical ventilation control. In P Ann Int IEEE EMBS 2016; 3626–3629. DOI: in Google Scholar

[5] Clancy EA, Hogan N. Probability density of the surface electromyogram and its relation to amplitude detectors. IEEE Trans Bio-Med Eng 1999; 46: 730–739.10.1109/10.764949Search in Google Scholar

[6] Comon P, Jutten C. Handbook of blind source separation: independent component analysis and applications. New York: Academic Press 2010.Search in Google Scholar

[7] De Luca CJ, Erim Z. Common drive of motor units in regulation of muscle force. Trends Neurosci 1994; 17: 299–305.10.1016/0166-2236(94)90064-7Search in Google Scholar

[8] De Luca CJ, Nawab SH, Kline JC. Clarification of methods used to validate surface EMG decomposition algorithms as described by Farina et al. (2014). J Appl Physiol 2015; 118: 1084.10.1152/japplphysiol.00061.2015Search in Google Scholar PubMed PubMed Central

[9] Deep K, Singh KP, Kansal ML, Mohan C. A real coded genetic algorithm for solving integer and mixed integer optimization problems. Appl Math Comput 2009; 212: 505–518.10.1016/j.amc.2009.02.044Search in Google Scholar

[10] Farina D, Enoka RM. Surface EMG decomposition requires an appropriate validation. J Neurophysiol 2011; 105: 981–982.10.1152/jn.00855.2010Search in Google Scholar PubMed

[11] Farina D, Merletti R. A novel approach for precise simulation of the EMG signal detected by surface electrodes. IEEE Trans Bio-Med Eng 2001; 48: 637–646.10.1109/10.923782Search in Google Scholar PubMed

[12] Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG. J Appl Physiol 2004; 96: 1486–1495.10.1152/japplphysiol.01070.2003Search in Google Scholar PubMed

[13] Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG: an update. J Appl Physiol 2014; 117: 1215–1230.10.1152/japplphysiol.00162.2014Search in Google Scholar

[14] Farina D, Merletti R, Enoka RM. Reply to De Luca, Nawab, and Kline: the proposed method to validate surface EMG signal decomposition remains problematic. J Appl Physiol 2015; 118: 1085.10.1152/japplphysiol.00107.2015Search in Google Scholar

[15] Fuglevand AJ, Winter DA, Patla AE. Models of recruitment and rate coding organization in motor-unit pools. J Neurophysiol 1993; 70: 2470–2488.10.1152/jn.1993.70.6.2470Search in Google Scholar

[16] Holobar A, Farina D. Blind source identification from the multichannel surface electromyogram. Physiol Meas 2014; 35: R143–R165.10.1088/0967-3334/35/7/R143Search in Google Scholar

[17] Hyvärinen A, Karhunen J, Oja E. editors. Independent component analysis. John Wiley & Sons, Inc.: New York 2001.10.1002/0471221317Search in Google Scholar

[18] Léouffre M, Quaine F, Servière C. Testing of instantaneity hypothesis for blind source separation of extensor indicis and extensor digiti minimi surface electromyograms. J Electromyogr Kines 2013; 23: 908–915.10.1016/j.jelekin.2013.03.009Search in Google Scholar

[19] Mambrito B, De Luca CJ. A technique for the detection, decomposition and analysis of the EMG signal. Electroen Clin Neuro 1984; 58: 175–188.10.1016/0013-4694(84)90031-2Search in Google Scholar

[20] Matsuoka K, Ohba Y, Toyota Y, Nakashima S. Blind separation for convolutive mixture of many voices. In Proc IWAENC 2003, 2003; 279–282.Search in Google Scholar

[21] McSharry PE, Clifford GD, Tarassenko L, Smith LA. A dynamical model for generating synthetic electrocardiogram signals. IEEE Trans Bio-Med Eng 2003; 50: 289–294.10.1109/TBME.2003.808805Search in Google Scholar PubMed

[22] Nawab SH, Chang S, De Luca CJ. High-yield decomposition of surface EMG signals. Clin Neurophysiol 2010; 121: 1602–1615.10.1016/j.clinph.2009.11.092Search in Google Scholar PubMed PubMed Central

[23] Negro F, Muceli S, Castronovo AM, Holobar A, Farina D. Multi-channel intramuscular and surface EMG decomposition by convolutive blind source separation. J Neural Eng 2016; 13: 026027.10.1088/1741-2560/13/2/026027Search in Google Scholar PubMed

[24] O’Brien MJ, van Eykern LA, Prechtl HFR. Monitoring respiratory activity in infants – a non-intrusive diaphragm EMG technique. In Rolfe, editor, Non-invasive measurements 2. Academic Press Inc. (London) Ltd. 1983.Search in Google Scholar

[25] Piquilloud L, Jolliet P. Comparative effects on diaphragmatic electrical activity and respiratory pattern of various levels of assistance, 2012. Unpublished study protocol, approved by the local ethics committee at Lausanne University Hospital (identifier: NCT01248845).Search in Google Scholar

[26] R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2015. URL in Google Scholar

[27] Sameni R. Open source ECG toolbox, 2006. URL Released under the GNU General Public License.Search in Google Scholar

[28] Sameni R, Clifford G, Jutten C, Shamsollahi M. Multichannel ECG and noise modeling: Application to maternal and fetal ECG signals. EURASIP J Adv Sig Pr 2007; 2007: Article id 043407, 14 pages. doi:10.1155/2007/4340710.1155/2007/43407Search in Google Scholar

[29] Sinderby C. Proportional pressure assist ventilation controlled by a diaphragm electromyographic signal, 2001. Patent.Search in Google Scholar

[30] Sinderby C, Navalesi P, Beck J, et al. Neural control of mechanical ventilation in respiratory failure. Nat Med 1999; 5: 1433–1436.10.1038/71012Search in Google Scholar PubMed

[31] Willigenburg NW, Daffertshofer A, Kingma I, van Dieën JH. Removing ECG contamination from EMG recordings: a comparison of ICA-based and other filtering procedures. J Electromyogr Kines 2012; 22: 485–493.10.1016/j.jelekin.2012.01.001Search in Google Scholar PubMed

Received: 2016-4-12
Accepted: 2016-11-28
Published Online: 2017-1-11
Published in Print: 2017-4-1

©2017 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 28.5.2023 from
Scroll to top button