Available via license: CC BY 4.0
Content may be subject to copyright.
Journal of Neural Engineering
PAPER • OPEN ACCESS
Adaptive HD-sEMG decomposition: towards
robust real-time decoding of neural drive
To cite this article: Dennis Yeung
et al
2024
J. Neural Eng.
21 026012
View the article online for updates and enhancements.
You may also like
Elbow-flexion force estimation during arm
posture dynamically changing between
pronation and supination
Ruochen Hu, Xiang Chen, Chengjun
Huang et al.
-
Using the Robust High Density-surface
Electromyography Features for Real-Time
Hand Gestures Classification
Hanadi A Jaber, Mofeed T Rashid and L
Fortuna
-
A robust HD-sEMG sensor suitable for
convenient acquisition of muscle activity in
clinical post-stroke dysphagia
Nan Zhao, Bolun Zhao, Gencai Shen et al.
-
This content was downloaded from IP address 188.94.86.74 on 13/04/2024 at 21:10
J. Neural Eng. 21 (2024) 026012 https://doi.org/10.1088/1741-2552/ad33b0
Journal of Neural Engineering
OPEN ACCESS
RECEIVED
19 September 2023
REVISED
21 February 2024
ACC EPT ED FOR PUB LICATI ON
13 March 2024
PUBLISHED
21 March 2024
Original Content from
this work may be used
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to
the author(s) and the title
of the work, journal
citation and DOI.
PAPER
Adaptive HD-sEMG decomposition: towards robust real-time
decoding of neural drive
Dennis Yeung1,∗, Francesco Negro2and Ivan Vujaklija1,∗
1Department of Electrical Engineering and Automation, Aalto University, Espoo, Finland
2Department of Clinical and Experimental Sciences, Universit`
a degli Studi di Brescia, Brescia, Italy
∗Authors to whom any correspondence should be addressed.
E-mail: dennis.yeung@aalto.fi and ivan.vujaklija@aalto.fi
Keywords: neural interfacing, decomposition, electromyography, human-machine interfacing, motor units
Abstract
Objective. Neural interfacing via decomposition of high-density surface electromyography
(HD-sEMG) should be robust to signal non-stationarities incurred by changes in joint pose and
contraction intensity. Approach. We present an adaptive real-time motor unit decoding algorithm
and test it on HD-sEMG collected from the extensor carpi radialis brevis during isometric
contractions over a range of wrist angles and contraction intensities. The performance of the
algorithm was verified using high-confidence benchmark decompositions derived from
concurrently recorded intramuscular electromyography. Main results. In trials where contraction
conditions between the initialization and testing data differed, the adaptive decoding algorithm
maintained significantly higher decoding accuracies when compared to static decoding methods.
Significance. Using “gold standard” verification techniques, we demonstrate the limitations of filter
re-use decoding methods and show the necessity of parameter adaptation to achieve robust neural
decoding.
1. Introduction
Force generation in human skeletal muscles is gov-
erned by the activity of constituent motor units
(MUs). Each MU is comprised of a single alpha motor
neuron and the set of muscle fibers that it innerv-
ates, where a single axonal action potential initiates
a tension-generating contractile twitch in the innerv-
ated fibers. The discharge pattern of a MU popula-
tion thus encode the neural drive underlying gross
muscular contraction [1,2]. Historically, the pre-
cise activation times of individual MUs were only
attainable via manual or semi-automatic spike sort-
ing of electromyography (EMG) signals measured
from indwelling electrodes [1,3–6]. More recently,
convolutive blind source separation techniques have
been developed to automatically extract MU spike
trains from high-density surface electromyography
(HD-sEMG) [7–9]. Such methods yield detailed
neural information in a non-invasive manner and
are capable of extracting far more MUs compared
to the spike sorting of intramuscular EMG (iEMG)
[10]. For these reasons, HD-sEMG decomposition
has garnered considerable interest in studies on
neurophysiology, motor control and neuromuscular
disorders [10–13]. In particular, MU decomposition
offers practical advantages over established modes of
human-machine interfacing (HMI) due to the access
to higher neural information without the need of
invasive procedures [14–16].
Traditionally, decomposition yields a set of sep-
aration vectors (MU filters) that distill HD-sEMG
into underlying source activities. This process relies
on repeated execution of iterative numerical meth-
ods over observations spanning substantial periods of
time, typically 10 s or more [8,17]. Hence, such batch
decomposition algorithms are unsuitable for real-
time deployment. Instead, reapplication of batch-
decomposed MU filters to real-time measurements
has been a commonly adopted approach [16,18,19].
However, these techniques assume surface MU action
potentials (sMUAPs) to remain consistent. In real-
ity, factors such as fatigue, contraction intensity, and
joint position alter the expression of sMUAPs on the
skin surface [20–23]. To tackle this challenge, decod-
ing algorithms that adapt to new data have been
© 2024 The Author(s). Published by IOP Publishing Ltd
J. Neural Eng. 21 (2024) 026012 D Yeung et al
developed [17,24]. However, these methods have
been tailored to specific conditions and are yet to
be evaluated against the gold standard reference of
iEMG-decomposed spike trains.
Here we propose a real-time MU decoding
algorithm that updates the MU filter and signal pre-
processing transforms as new action potentials of
the observed MU emerge. The algorithm was evalu-
ated on HD-sEMG recordings pertaining to isomet-
ric wrist extension contractions that vary across con-
traction intensities and joint angles. The accuracy of
the algorithm was verified using reference spike trains
manually decomposed from concurrently recorded
fine-wire iEMG.
2. Methods
2.1. HD-sEMG decomposition
The decomposition techniques employed in this work
are based on a convolutive mixture model for EMG
generation:
zi(k) =
J
∑
j=1
L−1
∑
l=0
aij (l)τj(k−l) + εi(k)(1)
where zi(k)is the value of HD-sEMG channel iat time
instant k.τj(k−l)is the pulse train of MU jwhile aij(l)
encodes its respective action potential. Lis therefore
the maximum duration of impulse responses that is
considered in the model and εi(k)is additive noise
inclusive of the activities of unextractable MUs.
The algorithm for batch decomposition is
described in detail in [8] though, here, a brief over-
view will be given for completeness. Firstly, the HD-
sEMG signals are extended by appending their time-
delayed versions as additional observations. This
conditions the data for the FastICA algorithm [25],
which normally decomposes instantaneous mixtures,
to handle convolutive mixtures [26]. Further pre-
processing of the observations includes zero-phase
component analysis (ZCA) sphering, which aids in
the convergence of FastICA [26]. The batch algorithm
then extracts underlying source activities in a sequen-
tial manner, thereby estimating the firing intervals of
MUs responsible for the generation of the observed
HD-sEMG. Each source signal is extracted as:
ˆ
s(k) = b⊤Σ−1/2
˜
z˜
z(˜
z(k)−µ˜
z)(2)
where ˜
z(k)is the extended observation vector and
µ˜
z=E[˜
z(k)] is the vector of subtractive means
for centering ˜
z(k).Σ−1/2
˜
z˜
zis the sphering mat-
rix, calculated as the inverse square root of the
covariance matrix of extended obsrvations, Σ˜
z˜
z=
E[(˜
z(k)−µ˜
z)(˜
z(k)−µ˜
z)⊺]. Finally, bis the spati-
otemporal filter that extracts the MU contribution.
The processes involved for simultaneously estim-
ating band ˆ
s(k)in a blind manner can be broken
down into the Extraction and Refinement step. The
Extraction step employs FastICA which iterates a
fixed-point algorithm with an objective function
optimizing the sparsity of ˆ
s(k). Orthogonalization of
MU filters is performed at every iteration to ensure
convergence to new sources. In the Refinement step,
the MU filters and spike trains are iteratively updated
to optimize the silhouette measure (SIL), a value
which measures the accuracy of the separation [8,23].
As per [7], each iteration first involves peak detection
on the estimated source signal. Spike classification is
then performed where the kmeans++ algorithm is
used to distinguish peaks as either spikes or noise,
with cluster centroids chi and clo, respectively. Finally
the MU filter is re-calculated as the cross-correlation
between the sphered, extended observations and the
current estimated spike train:
b=Σ−1/2
˜
z˜
z
1
|Ψ|∑
˜
zΨ∈Ψ
(˜
zΨ−µ˜
z)(3)
where ˜
zΨrepresents members in a set of extended
observations corresponding to time instants of estim-
ated spikes, Ψ={˜
z(k) : ˆτ(k) = 1}. The Refinement
step is thus repeated until the SIL value of the
re-estimated source ceases to increase and sources
with a final SIL value above a minimum acceptance
threshold are deemed as viable MU pulse trains.
2.2. Online decomposition
2.2.1. Static decoding
So far, the most prominent approach to estimat-
ing MU activities in real-time is through the re-
use of the MU filter, b, and pre-process transforms,
Σ−1/2
˜
z˜
zand µ˜
z, as presented by Barsakcioglu and
Farina [27]. These are initially obtained from train-
ing data and then continuously reapplied to new
windows of extended data in the same manner as
equation (2). Detected peaks in the estimated source
signal are further sorted as either spikes or noise
peaks. Rather than using the kmeans++ algorithm,
this is simply determined by a threshold set at the
midpoint between the spike and noise centroids, chi
and clo, also retained from batch decomposition of
the training data. To accommodate for deviations
between the conditions of the training data and the
new, unseen data, this decision boundary may be
altered by a relaxation factor, 0 ⩽α⩽1:
THR =clo + (1−α)chi −clo
2.(4)
2.2.2. Adaptive MU decoding
While relaxation of the decision boundary may
decrease the likelihood of missed spikes, it can also
lead to an increase in false positives. To address this,
we propose adaptation of the MU filter and prepro-
cess transforms as potential spike events are detec-
ted. Here, adaptation of decoding parameters occurs
in parallel to the online decoding process. The updat-
ing of the MU filter is performed in a manner similar
2
J. Neural Eng. 21 (2024) 026012 D Yeung et al
to the Refinement step of the batch algorithm. Since
this does not rely on FastICA, re-computation of ZCA
sphering transforms is unnecessary. Rather, only the
inverse of the sample covariance matrix is needed.
The equivalent of source extraction (equation (2))
for the adaptive decoding algorithm can therefore be
written as:
ˆ
s(k) = v⊤Σ−1
˜
z˜
z(˜
z(k)−µ˜
z)(5)
where vis now the MU filter and Σ−1
˜
z˜
zis the inverse
of the observation covariance matrix.
With each new data window, temporary trans-
forms are first derived from the updated statistics of
extended observations:
µ∗
˜
z=λµ˜
z+ (1−λ)E[˜
zwin (k)] (6)
Σ∗
˜
z˜
z=λΣ˜
z˜
z
+ (1−λ)E[(˜
zwin (k)−µ∗
˜
z)(˜
zwin (k)−µ∗
˜
z)⊺]
(7)
where ˜
zwin (k)is the kth sample in the new window of
extended data and 0 ⩽λ⩽1 is the forgetting factor
that controls the influence of new data. An initial
estimation of the source signal is then calculated:
ˆ
swin (k) = v⊤Σ∗−1
˜
z˜
z(˜
zwin (k)−µ∗
˜
z)(8)
where Σ∗−1
˜
z˜
zis the inverse of the temporary covari-
ance matrix obtained from equation (7).
Peak detection is then conducted on the estimated
source signal of the tracked MU, ˆ
swin(k). To identify
potential new spike instances for learning, the sparse-
ness property of MU firing patterns can be leveraged.
With a-priori knowledge regarding the short time-
span that the data window corresponds to, any strong
responses to the MU filter, meaning potential spikes,
will appear as outliers in the distribution of rectified
peak amplitudes. Hence, rectified peak amplitudes
with z-scores above a certain threshold will have their
corresponding extended observation vectors added
to set Ψmem. Functionally, Ψmem is implemented as a
first-in-first-out storage buffer of constant size, ini-
tialized from extended observations corresponding to
spike events in the training data. As candidate spikes
are detected from new data windows and new obser-
vation vectors are added to Ψmem, past observations
are discarded. With each update of Ψmem, the MU fil-
ter is recalculated using equation (9):
v=1
|Ψmem|∑
˜
zΨ∈Ψmem
(˜
zΨ−µ∗
˜
z).(9)
Equation (9), similar to equation (3), updates the
MU filter via cross-correlation between new extended
observations and the |Ψmem|most recently estimated
spikes.
The spike and noise centroids, which are used
for online spike detection outside of the adapta-
tion algorithm, are subsequently updated. The spike
centroid is recalculated via equation (10) which cor-
responds to the squared average amplitude of peaks
extracted from the observations stored in Ψmem. The
noise centroid is then updated as a λ-weighted mer-
ging of the past clo and the average of noise peak
amplitudes detected from the new source signal
window:
chi =(v⊺Σ∗−1
˜
z˜
zv)2(10)
clo =λclo
+ (1−λ)
v⊺Σ∗−1
˜
z˜
z
1
|ηwin|∑
˜
zη∈ηwin
(˜
zη−µ∗
˜
z)
2
(11)
where ηwin is the set of observations corresponding to
noise peaks detected inˆ
swin (i.e. peak observations not
added to Ψmem).
Algorithm 1summarizes this entire process for
real-time updating of decomposition parameters.
Prior to implementation, there are three static para-
meters that need to be defined: the threshold z-score
value for accepting new observations into Ψmem, the
rate of forgetting, λ, and the cardinality of Ψmem.
As in past studies, such parameters were selected
empirically [17,24]. For the results obtained in this
work, the corresponding values used were 3.3, 0.985
and 110, respectively, based on initial testing.
Algorithm 1. Adaptation of online MU decoding
parameters
1. Calculate µ∗with equation (6).
2. Calculate Σ∗
˜
z˜
zwith equation (7).
3. Calculate ˆ
swin(k)with equation (8).
4. Perform peak detection onˆ
swin(k).
5. Sort peak as spikes or noise based on z-scores of their
rectified amplitudes.
6. if Spike peaks detected then
7. Add new spike observations to Ψmem while
discarding an equal number of oldest
observations.
8. Build ηwin from observations corresponding to
noise peaks extracted from new data.
9. Calculate the new MU filter, v, with equation (9).
10. Calculate new spike and noise centroids with
equations (10) and (11).
11. Accept updated inverse covariance matrix:
Σ−1
˜
z˜
z←Σ∗−1
˜
z˜
z.
12. Accept updated statistics: µ˜
z←µ∗
˜
z,Σ˜
z˜
z←Σ∗
˜
z˜
z
13. end if
2.3. Experimental setup
Five able-bodied subjects were recruited for the
experiment, four male, one female, ages 29–34, all
right-handed. The study was approved by the local
ethical board of Aalto University (approval number
D/505/03.04/2022). Prior to the experiments, all sub-
jects gave their written informed consent in accord-
ance with the Declaration of Helsinki.
3
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Figure 1. Experimental setup. (a) Three fine-wire electrode pairs inserted into a subject’s ECRB. (b) A 64-channel high-density
surface electrode grid placed above the iEMG insertion sites shown in (a). (c) Subject with iEMG and HD-sEMG electrodes
attached to their dominant arm which has been placed inside the force measurement rig. Task cues are shown on the computer
screen in front of the subject.
Subjects were seated for the duration of the exper-
iment with their dominant upper-limb placed in a
specialized tabletop rig designed to constrain the
wrist joint at various angles of extension (figure 1).
Forces generated by isometric contractions pertain-
ing to wrist extension were measured with a load
cell (TAS606, HT Sensor Technology, China) at a
sampling rate of 100 Hz. Prior to the insertion of
fine-wire electrodes, the subject’s maximum volun-
tary contraction (MVC) forces were measured at wrist
joint angles corresponding to 0%, 12.5% and 25% of
their maximal extension, with 0% relating to a neut-
ral wrist position. MVC was calculated as the aver-
aged maximal force from three MVC contractions of
1.5 ms long with short breaks in between each con-
traction to prevent fatigue.
Stainless steel/silver (SS/Ag) wires with polytet-
rafluoroethylene insulation (Spes Medica s.r.l, Italy)
were used as intramuscular electrodes. The wires had
a diameter of 0.11 mm with the final 3–5 mm of
the recording tips stripped of the insulating mater-
ial. Three insertion points were targeted, centered at
the bulk of the extensor carpi radialis brevis (ECRB)
and aligned down the muscle axis at approximately
4 mm intervals. Location of the ECRB was guided by
[28] and palpation during wrist extension and radial
deviation movements. The fine-wires were inserted
as pairs (bipolar configuration) using 25G cannu-
lae to a depth targeting MUs proximal to the skin
surface. Signal inspection was conducted after the
insertion of each electrode pair. If the signal was
invalid (short-circuited, excessive noise, low selectiv-
ity or no viable units detected) and could not be
remedied by light manipulation of the fine-wires,
the wires were removed and another insertion of
new electrodes was made slightly lateral to the ori-
ginal insertion point. The maximum overall num-
ber of insertion attempts was bounded to five for the
sake of subject comfort. The experiment only pro-
ceeded so long as at least one valid iEMG channel
was attained. The bipolar iEMG signals were pre-
amplified by an adapter (ADx5JN, OT Bioelettronica,
Italy) with a gain of 5, and acquired by a bioamplifier
(Quattrocento, OT Bioelettronica, Italy) with a fixed
gain of 150 at 10 240 Hz with 10-4400 Hz hardware
bandpass filtering. Subsequent processing of iEMG
signals included high-pass filtering with a 250 Hz cut-
off to lower baseline noise and to produce sharper
action potentials [5,29].
Placement of the overlaying HD-sEMG matrix
was conducted approximately 8 minutes after the
final fine-wire insertion. This allowed for sufficient
coagulation, minimizing the leakage of blood or
plasma to the surface recording site. A 64-channel
rectangular electrode matrix (GR08MM1305, OT
Bioelettronica, Italy) with 8 mm inter-electrode dis-
tance was placed on top of the ECRB, centered above
the iEMG insertion sites (figure 1). Two reference
electrodes (Neuroline 720, Ambu A/S, Denmark),
one for the pre-amplifier and one for the bioamplifier,
were placed at the medial epicondyle and olecranon
process. The HD-sEMG signals were buffered by a
pre-amplifier (AD64F, OT Bioelettronica, Italy) prior
to being acquired by the same benchtop ampli-
fier used for iEMG at 150 gain, 10 240 Hz with 10-
4400 Hz hardware bandpass filtering. Pre-processing
of the HD-sEMG signals for automatic decomposi-
tion included downsampling to 2048 Hz and band-
pass filtering with 10-900 Hz cut-offs.
Prior to the commencement of recordings, sub-
jects were asked to perform slow dynamic wrist exten-
sions, up to 25% of maximum range of movement,
to allow the settling-in of fine-wire electrodes and
HD-sEMG matrix. The recording and cueing of con-
tractions were facilitated by a custom Matlab R2021b
(MathWorks Inc. USA) framework. All subject cues,
along with the real-time force feedback, were dis-
played on a computer screen.
2.4. Experimental protocol
Isometric wrist extension contractions with
trapezoidal force profiles (5 s ramp, 20 s plateau)
were recorded at different joint angles and different
force levels. To ensure iEMG decomposability, which
relies on low to moderate signal complexity [4,5],
contraction intensities were kept at low levels. For
4
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Figure 2. iMUAPs and sMUAPs extracted from different contraction conditions. Up to 3 unique iMUAP shapes are utilized for
manual matching of MUs extracted from different contraction conditions. While each individual iMUAP profile is still susceptible
to change across the angle and force conditions, causing potential matching ambiguities, the presence of multiple time-locked and
distinct profiles facilitate matching of MUs that may share similar iMUAP profiles in one particular channel. Variation in the
sMUAP profiles across contraction conditions is also observed, resulting in sub-optimal extraction of source activities when using
static decoding algorithms. (a) iMUAPs and sMUAPs of MU B1 extracted from all 5 contraction conditions that it was detected
in. From darkest to lightest plot lines, the displayed MUAPs correspond to angle/force combinations of 0%/5%, 0%/10%,
25%/5%, 25%/10% and 25%/15%, respectively. (b) iMUAPs and sMUAPs of MU B2 obtained from the same contraction
conditions as displayed in (a). (c) iMUAPs and sMUAPs of MU C1 extracted from 3 different contraction conditions. From
darkest to lightest plot lines, the displayed MUAPs correspond to angle/force combinations of 0%/5%, 12.5%/7.5% and
25%/10%, respectively. (d) iMUAPs and sMUAPs of MU C2 obtained from the same contraction conditions as displayed in (c).
subjects A and B, contractions were recorded at force
levels of 5%, 10% and 15% MVC at wrist joint angles
of 0% and 25% maximal extension. For subjects C,
D and E, contractions of 5%, 7.5% and 10% MVC
were recorded at 0%, 12.5% and 25% maximal wrist
extension. Recordings progressed from 0% to 25%
extension while the order of force levels recorded was
randomized. Three repetitions were recorded for each
contraction condition.
2.5. Obtaining iEMG decomposition benchmarks
2.5.1. Extraction of MU activity concurrent in iEMG
and sEMG
To identify MUs present in both surface and intra-
muscular signals, a two-stage semi-automatic tech-
nique was employed. For each repetition, a set of
MUs and their respective spike trains are first extrac-
ted from HD-sEMG via the batch decomposition
method described in section 2.1. The resultant spike
intervals were then used to trigger action poten-
tials in the iEMG signals. Here, MUs whose activities
are present in both the concurrently recorded HD-
sEMG and iEMG signals will trigger distinct intra-
muscular motor unit action potentials (iMUAPs).
Typically, these are mono and polyphasic waveforms
with peaks well above the baseline noise [4]. On
the other hand, MUs that were only extractable via
HD-sEMG decomposition will trigger flat iMUAPs
(peak-to-peak amplitudes <2µV). In this way, units
present in both surface and intramuscular record-
ings are identified. The spike-trains of such MUs
were then imported to EMGlab [29], a Matlab-based
spike annotation software, for manual correction by
an experienced operator such that a high-confidence
benchmark is obtained.
2.5.2. Tracking MUs across contraction conditions
MUs were matched by the same experienced operator
through visual comparison of their multi-channel
iMUAPs. As each iEMG channel consisted of a bipolar
measurement, activity from a single source mani-
fests as action potentials that vary greatly in profile
across channels but are, albeit, time-locked. Thus, a
single MU may be characterized by up to three dis-
tinct action potentials triggered by the same spike
instances. Examples are shown in figure 2where such
iMUAP profiles may be used to manually match MUs
across contraction conditions.
5
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Table 1. Catalog of MUs and trial conditions used for this study. 18 MUs were found to be identifiable, both in HD-sEMG and iEMG, in
at least two levels of joint angle and force. Hence, they satisfied the inclusion criterion for this study. MUs sharing the same letter in their
designation were extracted from the same subject and set of recordings.
Angle (%) Force (%)
MU
A1 A2 B1 B2 B3 C1 C2 C3 C4 C5 D1 D2 E1 E2 E3 E4 E5 E6
0
5o o o o o o o o o o o x o o o o o x
7.5 - - - - - o o o o o o x o o o o o o
10 o o o o o o o o x o o x o o o x x o
15 o x x x x - - - - - - - - - - - - -
12.5
5 - - - - - o o o o o x x o o o o o o
7.5 - - - - - o o o o o o o o o o o o o
10 - - - - - o o o o o o o o o o o o o
15 - - - - - - - - - - - - - - - - - -
25
5 o o o o o o o o o x x x o o x o o x
7.5 - - - - - o o o o x o o o o o o o o
10 o o o o o o o x o x o o o o o o o o
15 o o o o o - - - - - - - - - - - - -
o=MU concurrently detected in HD-sEMG and iEMG, x =MU not concurrently detected, - =trial not recorded for subject.
2.6. Pseudo-online testing
In the pseudo-online tests, multiple trials were con-
ducted to gauge the robustness of the proposed adapt-
ive MU decoding algorithm across different contrac-
tion conditions. In each trial, the decoding algorithm
was initialized from one repetition and then applied
to extract MU activity in another. Here, data was
fed in windows of 200 ms and in time increments of
100 ms, thereby simulating real-time deployment. For
comparative purposes, the static decoding technique
(section 2.2.1) was also tested using different spike
threshold relaxation values, from α=0 to α=0.5 in
increments of 0.1.
Since the decoding algorithms were to be com-
pared in scenarios where the conditions of the
training data differed from those of the test data,
only MUs with high-confidence iEMG-decomposed
benchmarks (obtained by methods described in
sections 2.5.1 and 2.5.2) in at least two force levels
for at least two angle conditions were selected for this
analysis. Table 1lists the MUs selected for this testing
along with the contraction conditions in which they
were detected in. For each eligible MU, all pair-wise
combinations of training and testing repetitions were
analyzed.
For each trial, the estimated spike train was com-
pared to the iEMG-decomposed spike train using the
Rate-of-Agreement (RoA) metric:
RoA =C
C+O+I·100%(12)
where Ois the number of spikes that were only detec-
ted by the online decomposition algorithm and Iis
the number of firing instances exclusive to the iEMG
decomposition while Cis the number of spikes that
were identified in both estimations of MU activity.
In addition, two metrics that are analogous to False
Negative Rate (FNR) and False Discovery Rate (FDR),
when considering the iEMG-decomposed spike train
as ground truth, were calculated for each trial:
FNR =I
I+C·100%(13)
FDR =O
O+C·100%.(14)
2.7. Statistical analysis
To detect statistically significant differences between
the decoder performances, repeated-measures ana-
lysis of variance (RM-ANOVA) was conducted on the
RoA values obtained from pseudo-online testing. The
normality of the results was verified with Shapiro-
Wilks’s testing while the assumption of sphericity
was tested with Mauchly’s test. In cases where the
sphericity assumption was not satisfied, Greenhouse-
Geisser correction was applied to the RM-ANOVA.
If the choice of decoder was found to have a signi-
ficant effect, post-hoc pair-wise comparisons (Tukey-
Kramer) were conducted. In all analyses, significance
levels of 0.05 were used.
In addition to analyzing the full set of res-
ults, three auxiliary analyses, ‘Intra-condition’, ‘Inter-
angle’ and ‘Inter-force’, were conducted on different
subsets of the data. In the Intra-condition analysis,
only the trials where the training and testing data had
identical angle/force conditions were considered. In
the Inter-angle analysis, trials where the training and
testing data were recorded from identical force con-
ditions, but had different angle conditions, were con-
sidered. Finally, the Inter-force analysis focused on
trials where the training and test recordings consisted
of contractions with identical angle but different force
conditions.
3. Results
Eighteen MUs (table 1) were found to satisfy the
inclusion criterion for this study. This yielded 6073
6
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Figure 3. (Top) Performance of static and adaptive MU decoding methods (SD and AD, respectively) in terms of RoA with
iEMG-referenced benchmark decompositions. Error bars indicate standard deviation. The Global analysis encompasses all
conducted trials (N=6073). The Intra-condition analysis (N=831) considers only trials in which contraction conditions of the
training repetition used for decoder initialization matched those of the test repetition. In the Inter-angle analysis (N=1192), only
trials with differing joint angle conditions, but identical force levels, between the training and test repetitions are included. For the
Inter-force analysis (N=1394), the included trials feature identical joint angle levels between the training and test repetitions but
differ in force level. RM-ANOVA was conducted on the RoA results with the decoding algorithm selected as the independent
variable. Statistically significant differences between the online decoders was found in all the analyses. Pairwise comparisons
between decoders with statically significant differences are indicated by ‘∗’ in the grids above the bar-charts. The proposed
adaptive algorithm is shown to achieve superior robustness as high RoA is maintained when tested contraction conditions differ
from those used for decoder initialization. (Middle and bottom) FNR and FDR for each the decoder is also shown. Adaptive
decoding is shown to negate the trade-off in increasing α, which lowers the occurrence of false negatives at the expense of higher
false positive rates.
trials in total, of which, 831, 1192, and 1394 were
included in the Intra-condition, Inter-angle, and
Inter-force analyses, respectively. Figure 3shows
the results from pseudo-online testing of the static
and adaptive decoders in terms of RoA with iEMG-
referenced benchmark decompositions. Statistically
significant differences between decoder performances
were detected in all analyses (F(1.99,1846.33) =
325.95, p<0.001, F(2.16,265.67) = 101.88,
p<0.001, F(1.99,356.77) = 87.99, p<0.001,
F(2.00,438.80) = 78.94, p<0.001 for Global, Intra-
condition, Inter-angle, and Inter-force, respect-
ively). In post-hoc comparisons, the proposed adapt-
ive decoding algorithm significantly outperformed
static decoding for all tested αvalues (0–0.5) in
the Global, Inter-angle, and Inter-force analyses
(p<0.001 for all comparisons). Overall, α=0.3 gave
the best static decoder performance with an RoA of
77.1%±25.2%in the Global analysis. Still, this was
exceeded by the adaptive decoder by 6.7%±0.2%.
Similarly, in the Inter-angle and Inter-force analyses,
the best static decoding performances were 75.7%±
25.0%(α=0.3) and 80.8%±22.4%(α=0.2) RoA,
respectively. The adaptive decoder also outperformed
these by 8.0%±0.4%and 5.1%±0.4%, respect-
ively. In the Intra-condition analysis, static decoding
is shown to still perform well with α=0.1 and 0.2
yielding the highest average RoAs of 94.1%±7.5%
and 93.9%±7.5%, respectively. Adaptive decoding
marginally underperformed these by −0.7%±0.2%
and −0.6%±0.2%, respectively.
Figure 3also shows decoding performances in
terms of FNR and FDR. Here, the effect of increasing
αis clearly shown. By relaxing the spike amplitude
threshold, fewer spikes are missed (lower FNR) but
in turn, more noise peaks are misclassified as spikes
(higher FDR). In contrast, the adaptive decoding
algorithm maintains low rates of either misclassifica-
tion types. Compared to static decoding with α=0.3,
which yielded 14.2%±13.7%FNR and 16.6%±
21.8%FDR in the Global analysis, adaptive decod-
ing achieved lower misclassifications by −0.5%±
0.2%and −8.1%±0.2%, respectively. The adaptive
decoding algorithm therefore resolves this trade-off
between FNR and FDR. Figure 4shows how this is
achieved by comparing the source activities extracted
7
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Figure 4. Estimated source signals and spike trains of MU E1 using static (top row) and adaptive (bottom row) decoders. Only
extraction of the first 4 s out of the full 12 s recordings are shown. Estimated spikes that agree with the corresponding
iEMG-referenced decomposition are indicated by green circles. Spike estimates that disagree are indicated by red x’s while missed
spikes are indicated by purple crosses. For static decoders, spike estimations using α=0 and 0.3 are shown. Spike amplitude
thresholds are plotted as black dotted lines and dash-dot lines for the relaxed threshold (α=0.3). The plots on the left side show
the application of decoding parameters initialized from a wrist angle/force level combination of 0%/7% on a contraction of
25%/10%. The right side plots show results with swapped initialization and test data. With static decoding, the extracted signals
are noisy and result in numerous misclassifications. The amount of missed spikes can be reduced by relaxing the spike amplitude
threshold but this results in higher occurrences of misidentified spikes. With adaptive decoding, continuous updating of
decomposition parameters maintain a clear separation of spike and noise peaks which result in higher decoding accuracies.
Figure 5. RoA results from tracking MU E1 using static and adaptive decoders. All possible pairwise combinations of training and
testing repetition are shown. Cells close to the heatmap diagonals represent results used for the ‘Intra-condition’ analysis as the
decoders are tested on data pertaining to the same contraction conditions used for initialization. In such cases, static decoding
with no relaxation of the spike detection threshold (α=0, left) maintains high RoA with iEMG-referenced benchmarks.
However, when tracking MU activity in contraction conditions that differ from the training data, RoA decreases. In some cases,
this can be remedied by increasing α(middle), though the majority of decoding accuracies remain poor. The proposed adaptive
decoder (right) therefore offers the best robustness with the majority of trials yielding RoA above 90%. However, the algorithm
may not always compensate for some large changes in contraction conditions as shown by the few trials with low RoA (<50%)
results. For instance, the algorithm can fail to converge to an appropriate filter when presented with contractions of wrist
angle/force level combinations of 25%/10% when using decoding parameters initialized from contractions of 0%/5%.
via online decoding with static and adapting paramet-
ers. By updating the MU filtering parameters as new
data is received, a clear separation of spike peaks and
noise peaks in the extracted source is maintained.
The RoA values of all trials regarding a single
MU are shown in figure 5. Results obtained via
static decoding with α=0 and 0.3, which yiel-
ded the best overall static decoding performance,
and the proposed adaptive decoding algorithm are
shown. The adaptive algorithm yielded similar decod-
ing accuracies as static decoders in the trials that
fall under the Intra-condition analysis (cells prox-
imal to the diagonals of the heatmaps). However,
in the majority of trials where the training con-
dition does not match the test condition, adapt-
ive decoding achieved a much higher RoA with the
iEMG-decomposed benchmarks. Still, there remain
cases where adaptation is unable to compensate for
the large changes to the sMUAPs.
4. Discussions
We have proposed an adaptive algorithm for decod-
ing MU activity from HD-sEMG that continuously
updates its internal parameters in real-time, as new
measurements are acquired. Using experimental data,
we demonstrate the performance of our proposed
algorithm in tracking MU activities across isomet-
ric contraction conditions that vary in joint angle
and intensity. In comparison to the static, non-
adaptive decoding algorithm, adaptive decoding was
shown to be more robust to such changes. This
8
J. Neural Eng. 21 (2024) 026012 D Yeung et al
was verified against benchmark spike trains manu-
ally decomposed from iEMG signals that were recor-
ded concurrently with the HD-sEMG. In terms of
RoA between decoder estimations and the iEMG-
referenced benchmarks, the adaptive decoder signi-
ficantly outperformed static decoders across all tested
spike threshold relaxation values (figure 3). Even
when the test trials only differed from the training
trial by one factor (Inter-angle or Inter-force), adapt-
ive decoding was shown to be beneficial. Nonetheless,
static decoding was still effective in estimating MU
activity from contractions that are similar to the train-
ing data (Intra-condition).
As contraction intensity and joint angle change,
so do the sMUAP profiles (figure 2). This renders
sphering transforms and MU filters derived from
different contraction conditions to be sub-optimal
for accurate source estimation [30,31], resulting in
missed spikes. In [23], local batch optimization of
MU filters allowed for the accurate decomposition of
MU activity during dynamic contractions. However,
this required prior knowledge regarding the period-
icity of the dynamic contractions. Here, we demon-
strate that relaxation of spike acceptance thresholds
can help compensate for changes to sMUAPs but
this also causes an increase in false spike identi-
fications, as evidenced by the inverse relationship
between the FNR and FDR results obtained using
static decoding (figure 3). Conversely, the adaptive
decoder maintains low rates of both false negative
and false positive errors. This is achieved by adapt-
ation of pre-process transforms and the MU filter
as new spikes are estimated, which helps maintain
a distinct separation between spike and noise peaks
(figure 4).
Previous studies on adaptive, real-time decod-
ing algorithms include [17] and [24], both of which
are based on the convolutional kernel compensa-
tion algorithm [7]. In [17], an adaptive decoding
algorithm was tested on a set of isometric contrac-
tions, ranging from 5% to 20% MVC, that were
recorded from the tibialis anterior of eight subjects.
Compared to spike trains extracted via batch decom-
position, the real-time algorithm achieved an average
RoA of 83%. In this work, we have achieved com-
parable accuracies (figure 3) in the more challenging
scenario of decoding across contraction conditions.
In [24], a separate algorithm was applied to decode
simulated and experimental dynamic contractions.
While dynamic contractions better represent the user
input of HMI applications, only pulse-to-noise ratio
was used to gauge decoder accuracy. In our study, the
proposed algorithm has been directly verified using
benchmark spike trains decomposed from intramus-
cular signals which remain the gold standard in the
field [32,33].
Beyond algorithmic adaptation, in [34], the
online decomposition algorithm was extended by
incorporating a self-administered enhancement
process utilizing the FitzHugh–Nagumo resonance
model. This extension aimed to enhance MU source
signals. When decomposing synthetic HD-sEMG
signals where MU recruitment, sMUAP amplitude,
and additive noise were varied, significant improve-
ment over the baseline decomposition algorithm was
achieved (88.70%±4.17%vs. 92.43%±2.79%).
Hence, employing physiologically-inspired models
for further signal processing, in conjunction with
adaptive decomposition, may yield even greater
decomposition robustness.
4.1. Limitations and future work
Currently, our proposed algorithm has only been
tested in a pseudo-online manner as verification
of decoding accuracy against manually decom-
posed benchmarks necessitates offline procedures.
Nonetheless, the algorithm is appropriate for real-
time deployment. In this study, data windows were
advanced in time increments of 100 ms while the
average execution time of parameter adaptation,
along with spike estimation, was 57.1±14 ms on
a desktop computer (Intel Xeon W-2133, 3.6 GHz, 36
GB RAM, Microsoft Windows 10, 64 bit). The main
computational cost to the algorithm lies in the com-
putation of the inverse covariance matrix. Despite
this, prior testing has shown that omission of this step
is detrimental to the overall effectiveness of the adapt-
ive decoding algorithm. One way to reduce compu-
tational demand is to reduce the extension factor. In
this work, we have used an extension factor of 16 to
align with established works [8,35]. However, past
investigations suggest that lower extension factors can
be employed and still retain user-intention estimation
performances in HMI applications [36]. While this
work focuses on verifying the accuracy of the pro-
posed adaptive decoding algorithm against iEMG-
referenced benchmarks, deployment in real-time
interfacing applications is left for future works.
Manual and semi-automatic decomposition of
iEMG by experienced operators has been extensively
verified from decades of research [4,37]. For this
reason, comparison with the decomposition of con-
currently recorded iEMG, sometimes referred to as a
variant of the two-source method [38–40], remains
the most convincing means of experimental valid-
ation with regards to the decomposition of surface
signals [32,38,41,42]. However, low signal com-
plexities are required to ensure the decomposabil-
ity of iEMG [4,5]. Hence, force levels were kept
to 15% MVC and below in this study. Indeed, past
works reliant on iEMG for verification have been sim-
ilarly constrained to low level contractions [17,38,
9
J. Neural Eng. 21 (2024) 026012 D Yeung et al
41,42]. While we have directly verified that our pro-
posed adaptive algorithm can compensate for sMUAP
changes at lower force levels, further verification at
higher force levels will have to depend on indirect
measures of decomposition accuracy or simulations
[24,31,34]. For instance, the application of static fil-
ter re-use in contractions of up to 70% MVC have
been previously investigated via such methods [31].
The issue of contraction intensity is closely linked
to the nature of action potential superposition.
The batch decomposition algorithm [8], which our
work extends upon, is based on blind source sep-
aration which inverts the mixing process modelled
by equation (1) and accounts for action poten-
tials from different sources overlapping in time and
space. The calculation of spatiotemporal filters cap-
able of extracting MU activities despite such over-
laps is aided by: 1) the high number of concurrent
observations afforded by the high-density electrode
grid, 2) statistical power afforded by longer record-
ings. The former aspect aids in discrimination of
sources with spatially overlapped action potentials
[43] while the latter permits computation of a spher-
ing transformation which decorrelates the extended
observations. This spatiotemporal decorrelation loc-
alizes sMUAPs in space and time, aiding in their
separation. However, at higher contraction intens-
ities, signal complexity from sMUAP superposition
can increase significantly due to heightened MU
recruitment and rate coding. Moreover, distinguish-
ing sMUAPs between unique sources becomes more
difficult due to the low-pass filtering effects of tissue
volume conduction [43]. This limitation is common
to many blind source separation-based decomposi-
tion algorithms [44] though it may be overcome with
signal acquisition methods that offer greater spatial
selectivity [43].
The adaptive decoding algorithm may not always
compensate for large, sudden changes in sMUAP pro-
files. As show in figure 5, when the difference between
the training and test contraction conditions is signi-
ficant, the adaptive algorithm may fail to converge
to the correct filtering parameters. Here, the decod-
ing algorithm is presented with an abrupt change
from one isometric contraction to another, whereas,
in practice, such changes occur in a continuous man-
ner. Hence, future works will also focus on the applic-
ation of adaptive decoding over experimental data
pertaining to dynamic contractions.
5. Conclusions
In conclusion, we have developed an adaptive MU
decoding algorithm that adapts to new data in
real-time. Using high-confidence in-vivo-referenced
benchmarks, the proposed algorithm was demon-
strated to be more accurate in decoding MU activities
across varying states of isometric contractions. This
work therefore paves the way towards robust, real-
time non-invasive neural interfacing.
Data availability statement
The data cannot be made publicly available upon
publication due to legal restrictions preventing
unrestricted public distribution. The data that sup-
port the findings of this study are available upon
reasonable request from the authors.
Acknowledgments
This work was supported by the Academy of Finland
Project No. 333149, ‘Hi-Fi BiNDIng’ (I V), and ERC
Consolidator Grant ID: 101045605, ‘INcEPTION’ (F
N). We acknowledge the computational resources
provided by the Aalto Science-IT project.
ORCID iDs
Dennis Yeung https://orcid.org/0000-0002-9760-
0752
Francesco Negro https://orcid.org/0000-0002-
9094-8932
Ivan Vujaklija https://orcid.org/0000-0002-7394-
9474
References
[1] Milner-Brown H S, Stein R B and Yemm R 1973 The
contractile properties of human motor units during
voluntary isometric contractions J. Physiol. 228 285
[2] Duchateau J and Roger M E 2011 Human motor unit
recordings: Origins and insight into the integrated motor
system Brain Res. 1409 42–61
[3] de Luca C J, LeFever R S, McCue M P and Xenakis A P 1982
Behaviour of human motor units in different muscles during
linearly varying contractions J. Physiol. 329 113
[4] Merletti R and Farina D 2008 Analysis of intramuscular
electromyogram signals Phil. Trans. R. Soc. A367 357–68
[5] Florestal J R, Mathieu P A and McGill K C 2009 Automatic
decomposition of multichannel intramuscular EMG signals
J. Electromyogr. Kinesiol. 19 1–9
[6] Marateb H R, Muceli S, McGill K C, Merletti R and Farina D
2011 Robust decomposition of single-channel intramuscular
EMG signals at low force levels J. Neural Eng. 8066015
[7] Holobar A and Zazula D 2007 Multichannel blind source
separation using convolution Kernel compensation IEEE
Trans. Signal Process. 55 4487–96
[8] Negro F, Muceli S, Margherita Castronovo A, Holobar A and
Farina D 2016 Multi-channel intramuscular and surface
EMG decomposition by convolutive blind source separation
J. Neural Eng. 13 026027
[9] Chen M and Zhou P 2016 A Novel Framework Based on
FastICA for High Density Surface EMG Decomposition
IEEE Trans. Neural Syst. Rehabil. Eng. 24 117–27
[10] Farina D, Negro F, Muceli S and Enoka R M 2016 Principles
of motor unit physiology evolve with advances in technology
Physiology 31 83–94
[11] Enoka R M and Farina D 2021 Force steadiness: from motor
units to voluntary actions Physiology 36 114–30
[12] Holobar A, Glaser V, Gallego J A, Dideriksen J L and
Farina D 2012 Non-invasive characterization of motor unit
behaviour in pathological tremor J. Neural Eng. 9056011
[13] Nishikawa Y, Holobar A, Watanabe K, Takahashi T, Ueno H,
Maeda N, Maruyama H, Tanaka S and Hyngstrom A S 2022
10
J. Neural Eng. 21 (2024) 026012 D Yeung et al
Detecting motor unit abnormalities in amyotrophic lateral
sclerosis using high-density surface EMG Clin. Neurophysiol.
142 262–72
[14] Farina D, Vujaklija I, Sartori M, Kapelner T, Negro F,
Jiang N, Bergmeister K, Andalib A, Principe J and
Aszmann O C 2017 Man/machine interface based on the
discharge timings of spinal motor neurons after targeted
muscle reinnervation Nat. Biomed. Eng. 125
[15] Holobar A and Farina D 2021 Noninvasive neural interfacing
with wearable muscle sensors: combining convolutive blind
source separation methods and deep learning techniques for
neural decoding IEEE Signal Process. Mag. 38 103–18
[16] Mendez Guerra I, Barsakcioglu D Y, Vujaklija I,
Wetmore D Z and Farina D 2022 Far-field electric potentials
provide access to the output from the spinal cord from
wrist-mounted sensors J. Neural Eng. 19 026031
[17] Glaser V, Holobar A and Zazula D 2013 Real-time motor
unit identification from high-density surface EMG IEEE
Trans. Neural Syst. Rehabil. Eng. 21 949–58
[18] Barsakcioglu D Y, Bracklein M, Holobar A and Farina D
2020 Control of spinal motoneurons by feedback from a
non-invasive real-time interface IEEE Trans. Biomed. Eng.
68 1
[19] Chen C, Yang Y, Sheng X, Farina D and Zhu X 2021
Simultaneous and proportional control of wrist and hand
movements by decoding motor unit discharges in real time J.
Neural Eng. 18 056010
[20] Farina D, Gazzoni M and Camelia F 2004 Low-threshold
motor unit membrane properties vary with contraction
intensity during sustained activation with surface EMG
visual feedback J. Appl. Physiol. 96 1505–15
[21] Pasquet B, Carpentier A and Duchateau J 2005 Change in
muscle fascicle length influences the recruitment and
discharge rate of motor units during isometric contractions
J. Neurophysiol. 94 3126–33
[22] Piitulainen H, Rantalainen T, Linnamo V, Komi P and
Avela J 2009 Innervation zone shift at different levels of
isometric contraction in the biceps brachii muscle J.
Electromyogr. Kinesiol. 19 667–75
[23] Souza Oliveira A and Negro F 2021 Neural control of
matched motor units during muscle shortening and
lengthening at increasing velocities J. Appl. Physiol.
130 1798–813
[24] Chen C, Shihan M, Sheng X, Farina D and Zhu X 2020
Adaptive real-time identification of motor unit discharges
from non-stationary high-density surface electromyographic
signals IEEE Trans. Biomed. Eng. 67 3501–9
[25] Hyvarinen A 1999 Fast and robust fixed-point algorithms for
independent component analysis IEEE Trans. Neural Netw.
10 626–34
[26] Thomas J, Deville Y and Hosseini S 2006 Time-domain fast
fixed-point algorithms for convolutive ICA IEEE Signal
Process. Lett. 13 228–31
[27] Barsakcioglu D Y and Farina D 2018 A real-time surface
EMG decomposition system for non-invasive
human-machine interfaces 2018 IEEE Biomedical Circuits
and Systems Conf., BioCAS 2018 - Proc.
[28] Leis A A and Trapani V C 2000 Atlas of Electromyography
(Oxford University Press)
[29] Kevin C M, Zoia C L and Marateb H R 2005 EMGLAB: an
interactive EMG decomposition program J. Neurosci.
Methods 149 121–33
[30] Glaser V and Holobar A 2019 Motor unit identification from
high-density surface electromyograms in repeated dynamic
muscle contractions IEEE Trans. Neural Syst. Rehabil. Eng.
27 66–75
[31] Francic A and Holobar A 2021 On the reuse of motor unit
filters in high density surface electromyograms recorded at
different contraction levels IEEE Access 9115227–36
[32] Farina D, Holobar A, Merletti R and Enoka R M 2010
Decoding the neural drive to muscles from the surface
electromyogram Clin. Neurophysiol. 121 1616–23
[33] Enoka R M 2019 Physiological validation of the
decomposition of surface EMG signals J. Electromyogr.
Kinesiol. 46 70–83
[34] Zheng Y, Guanghua X, Yixin Li and Qiang W 2022 Improved
online decomposition of non-stationary electromyogram via
signal enhancement using a neuron resonance model: a
simulation study J. Neural Eng. 19 026030
[35] Martinez-Valdes E, Negro F, Laine C M, Falla D, Mayer F and
Farina D 2017 Tracking motor units longitudinally across
experimental sessions with high-density surface
electromyography J. Physiol. 595 1479–96
[36] Yeung D, Negro F and Vujaklija I 2022 Effects of
decomposition parameters and estimator type on
pseudo-online motor unit based wrist joint angle prediction
ICNR 2020: Converging Clinical and Engineering Research on
Neurorehabilitation (Biosystems and Biorobotics) vol 28
pp 371–5
[37] Stashuk D 2001 EMG signal decomposition: how can
it be accomplished and used? J. Electromyogr. Kinesiol.
11 151–73
[38] Xiaogang H, Rymer W Z and Suresh N L 2014 Accuracy
assessment of a surface electromyogram decomposition
system in human first dorsal interosseus muscle J. Neural
Eng. 11 026007
[39] Holobar A, Minetto M A and Farina D 2014 Accurate
identification of motor unit discharge patterns from
high-density surface EMG and validation with a novel
signal-based performance metric J. Neural Eng.
11 016008
[40] Chen M, Zhang X, Zhiyuan L, Xiaoyan Li and Zhou P 2018
Two-source validation of progressive FastICA Peel-off for
automatic surface emg decomposition in human
first dorsal interosseous muscle Int. J. Neural Syst.
28 1850019
[41] Holobar A, Alessandro Minetto M, Botter A, Negro F and
Farina D 2010 Experimental analysis of accuracy in the
identification of motor unit spike trains from high-density
surface EMG IEEE Trans. Neural Syst. Rehabil. Eng.
18 221–9
[42] Kenneth Clarke A, Farokh Atashzar S, Del Vecchio A,
Barsakcioglu D, Muceli S, Bentley P, Urh F, Holobar A and
Farina D 2021 Deep learning for robust decomposition of
high-density surface EMG signals IEEE Trans. Biomed. Eng.
68 526–34
[43] Farina D, Negro F, Gazzoni M and Enoka R M 2008
Detecting the unique representation of motor-unit action
potentials in the surface electromyogram J. Neurophysiol.
100 1223–33
[44] Holobar A, Farina D, Gazzoni M, Merletti R and Zazula D
2009 Estimating motor unit discharge patterns from
high-density surface electromyogram Clin. Neurophysiol.
120 551–62
11
