Using a Support Vector Machine Algorithm to Classify Lower-Extremity EMG Signals During Running Shod/Unshod With Different Foot Strike Patterns

in Journal of Applied Biomechanics
Restricted access

Purchase article

USD  $24.95

Student 1 year subscription

USD  $87.00

1 year subscription

USD  $116.00

Student 2 year subscription

USD  $165.00

2 year subscription

USD  $215.00

The aim of this study is to use a support vector machine algorithm to identify and classify shod and barefoot running as well as rearfoot and forefoot landings. Ten habitually shod runners ran at self-selected speed. Thigh and leg muscle surface electromyography were recorded. Discrete wavelet transformation and fast Fourier transformation were used for the assembly of vectors for training and classification of a support vector machine. Using the fast Fourier transformation coefficients for the gastrocnemius and tibialis anterior muscles presented the best results for differentiating between rearfoot/forefoot running in the window before foot-floor contact possibly due to these muscles’ critical role in determining which part of the foot will first touch the floor. The classification rate was 76% and 67%, respectively, with a probability of being random of 0.5% and 4%, respectively. For the same terms and conditions of classification, the discrete wavelet transformation produced a reduction in the percentage of correctness of 60% and 53% with a probability of having reached these levels randomly of 15% and 35%. In conclusion, based on electromyographic signals, the use a fast Fourier transformation to train a support vector machine was a better option to differentiate running forefoot/rearfoot than to use the discrete wavelet transformation. Shod/barefoot running that could not be differentiated.

Pires, Falcari, and Campo are with the Laboratório de Controle Aplicado, Instituto Federal de Educação, Ciência e Tecnologia de São Paulo, São Paulo, SP, Brazil. Pulcineli and Ervilha are with the Escola de Artes, Ciências e Humanidades, Universidade de São Paulo, São Paulo, SP, Brazil. Hamill is with the Department of Kinesiology, University of Massachusetts Amherst, Amherst, MA, USA.

Ervilha (ulyervil@usp.br) is corresponding author.
  • 1.

    McCallion C, Donne B, Fleming N, Blanksby B. Acute differences in foot strike and spatiotemporal variables for shod, barefoot or minimalist male runners. J Sport Sci Med. 2014;13(2):280–286. PubMed ID: 24790480

    • Search Google Scholar
    • Export Citation
  • 2.

    Fredericks W, Swank S, Teisberg M, Hampton B, Ridpath L, Hanna JB. Lower extremity biomechanical relationships with different speeds in traditional, minimalist, and barefoot footwear. J Sport Sci Med. 2015;14(2):276–283. PubMed ID: 25983575

    • Search Google Scholar
    • Export Citation
  • 3.

    Von Tscharner V, Goepfert B, Nigg BM. Changes in EMG signals for the muscle tibialis anterior while running barefoot or with shoes resolved by non-linearly scaled wavelets. J Biomech. 2003;36(8):1169–1176. PubMed ID: 12831743 doi:10.1016/S0021-9290(03)00086-1

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 4.

    Wakeling JM, Von Tscharner V, Nigg BM, Stergiou P. Muscle activity in the leg is tuned in response to ground reaction forces. J Appl Physiol. 2001;91:1307–1317. PubMed ID: 11509530 doi:10.1152/jappl.2001.91.3.1307

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5.

    Komi P, Gollhofer A, Schmidtbleicher D, Frick U. Interaction between man and shoe in running: considerations for a more comprehensive measurement approach. Int J Sports Med. 1987;8(3):196–202. PubMed ID: 3623781 doi:10.1055/s-2008-1025655

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6.

    Hermens HJ, Merletti R, Rix H, Freriks B. The State of the Art on Signal Processing Methods for Surface Electromyography, Deliverable of the SENIAM project. Enschede, Netherlands: Roessingh Research and Development; 1999.

    • Search Google Scholar
    • Export Citation
  • 7.

    Ervilha UF, Mochizuki L, Figueira A, Hamill J. Are muscle activation patterns altered during shod and barefoot running with a forefoot footfall pattern? J Sports Sci. 2017;35(17):1697–1703. PubMed ID: 27626955 doi:10.1080/02640414.2016.1231410

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8.

    Bradski G. The OpenCV library. Dr Dobb’s J Softw Tools. 2000.

  • 9.

    Eaton JW, Bateman D, Hauberg S, Wehbring R. GNU Octave Version 3.8.1 Manual: A High-Level Interactive Language for Numerical Computations. CreateSpace Independent Publishing Platform; 2014. http://www.gnu.org/software/octave/doc/interpreter/

    • Search Google Scholar
    • Export Citation
  • 10.

    Galassi M, Davies J, Theiler J, et al. GNU Scientific Library Reference Manual. 3th ed. United Kingdom: Network Theory Ltd; 2009.

  • 11.

    Zhang X, Wang Y, Han RPS. Wavelet transform theory and its application in EMG signal processing. Paper presented at: 2010 Seventh International Conference on Fuzzy Systems and Knowledge Discovery; 2010. Yantai, China.

    • Export Citation
  • 12.

    Gunn SR. Support vector machines for classification and regression. ISIS Tech Rep. 1998;14(1):5–16.

  • 13.

    Yi Z, Li P, Zhu X, et al. Extracting time-frequency feature of single-channel vastus medialis EMG signals for knee exercise pattern recognition. PLoS ONE. 2017;12(7):1–18.

    • Search Google Scholar
    • Export Citation
  • 14.

    Stirling LM, von Tscharner V, Kugler PF, Nigg BM. Classification of muscle activity based on effort level during constant pace running. J Electromyogr Kinesiol. 2011;21(4):566–571. PubMed ID: 21459608 doi:10.1016/j.jelekin.2011.02.005

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 123 123 4
Full Text Views 9 9 1
PDF Downloads 2 2 0