Non-Negative Matrix Factorization of Simulated High Density Surface Electromyograms Reflects Both Muscle Excitation and Muscle Shortening

Abstract

We analyzed muscle excitation estimation systematically by Non-negative matrix factorization (NMF) from surface electromyograms (EMG) during dynamic contractions of biceps brachii (BB) muscles. We used motor unit action potentials (MUAPs) estimated experimentally from surface EMGs during slow dynamic contractions of BB muscles in healthy young males, and convolved them by simulated motor unit firing patterns. Different uncorrelated muscle excitation and muscle shortening profiles were combined when generating the EMG signals from left and right BBs (64 channels per muscle). EMG signals were rectified, low-passed filtered and decomposed by NMF into 2, 3, 4 or 6 components. The identified NMF components demonstrated good separation of left and right BB activity, but relatively large sensitivity of NMF components to muscle shortening, especially at high levels of muscle excitation. When averaged across different numbers of identified NMF components at excitation levels ranging from 40 to 80 %, the average correlation coefficient between the NMF components and muscle shortening profiles was 0.45 ± 0.15. At contraction levels between 0 and 40 % these correlations decreased to 0.15 ± 0.09. Therefore, NMF components reflect both muscle excitation and muscle shortening profiles.

Blind source separation of inspiration and expiration in respiratory sEMG signals

“…The thresholded wavelet bands d EMG i are further used to build the feature matrix V required for the following NMF algorithm [28]. Usually, in applications of NMF to EMG signals, the EMG envelope (as opposed to the raw signal) is used [4], [10], [12]- [16], [34], [36], [37]. Similarly, we calculate the envelopes of d EMG i in this approach by…”

“…Specifically, this work refines a method for suppressing cardiac activity [27] and interfaces the results to a BSS algorithm. The latter employs the NMF algorithm as a tool for identifying muscle synergies [4], [10]- [18], [34]- [37]. Here this work contributes in three ways.…”

Inter-Person Differences in Isometric Coactivations of Triceps Surae and Tibialis Anterior Decrease in Young, but Not in Older Adults After 14 Days of Bed Rest

“…Third, EMG amplitude envelopes (in bipolar or multichannel recordings) do not discriminate between the contributions of different MUs, and do not address the separation of contributions from different muscles when significant muscle crosstalk is present (for example, when recording activity of wrist extensors). Fourth, EMG amplitude envelopes are sensitive to changes of MUAPs that are caused either by contraction levels (Vieira et al, 2015), or by muscle shortening and muscle fatigue (Šavc and Holobar, 2021). At least at higher contraction levels, the fatigue may Abbreviations: CST, Cumulative Spike Train; EMG, Electromyography; GM, Gastrocnemius medialis; GL, Gastrocnemius lateralis; hdEMG, High-density Electromyography; MU, Motor Unit; MUAP, Motor Unit Action Potential; MVC, Maximum Voluntary Contraction; NMF, Non-negative Matrix Factorization; PNR, Pulse-to-Noise Ratio; RMS, Root-Mean-Square difference; SDR, Smoothed motor unit discharge rate; SO, Soleus; TA, Tibialis anterior.…”

“…At least at higher contraction levels, the fatigue may Abbreviations: CST, Cumulative Spike Train; EMG, Electromyography; GM, Gastrocnemius medialis; GL, Gastrocnemius lateralis; hdEMG, High-density Electromyography; MU, Motor Unit; MUAP, Motor Unit Action Potential; MVC, Maximum Voluntary Contraction; NMF, Non-negative Matrix Factorization; PNR, Pulse-to-Noise Ratio; RMS, Root-Mean-Square difference; SDR, Smoothed motor unit discharge rate; SO, Soleus; TA, Tibialis anterior. cause significant MUAP changes, decreasing the accuracy of muscle activation estimation (Holobar and Farina, 2021;Šavc and Holobar, 2021). Like MU territories and fibers' localizations, the fatiguing profiles and changes of MUAP shapes can be person specific.…”

“…Muscle activation patterns were also studied intensively by Non-negative Matrix Factorization (NMF) of the EMG amplitude envelopes (d 'Avella et al, 2003'Avella et al, , 2015Tresch et al, 2006). However, it was shown recently that NMF does not remove the MUAP shapes, and is, thus, sensitive to MU fiber localizations, changes of MUAP shapes due to muscle shortening or muscle fatigue (Šavc and Holobar, 2021), and to muscle crosstalk (Šavc et al, 2018). Therefore, also in the case of NMF-based analysis, the inter-person diversity of assessed muscle coactivation patterns could originate from person-specific MU properties.…”