sensors

Article
Validation of a Low-Cost Electromyography (EMG)
System via a Commercial and Accurate EMG Device:
Pilot Study
Sergio Fuentes del Toro 1,2, * , Yuyang Wei 3 , Ester Olmeda 1,2 , Lei Ren 3 , Wei Guowu 4 and
Vicente Díaz 1,2
1 Mechanical Engineering Department, Universidad Carlos III de Madrid, Avda. de la Universidad 30,
28911 Leganés, Spain; eolmeda@ing.uc3m.es (E.O.); vdiaz@ing.uc3m.es (V.D.)
2 Institute for Automotive Vehicle Safety (ISVA), Universidad Carlos III de Madrid,
Avda. de la Universidad 30, 28911 Leganés, Spain
3 School of Mechanical, Aerospace and Civil Engineering, University of Manchester, Manchester M13 9PL, UK;
yuyang.wei@manchester.ac.uk (Y.W.); lei.ren@manchester.ac.uk (L.R.)
4 School of Science, Engineering and Environment, University of Salford, Salford M5 4WT, UK;
G.Wei@salford.ac.uk
* Correspondence: sfuentes@ing.uc3m.es; Tel.: +34-916-624-9912




Received: 29 October 2019; Accepted: 26 November 2019; Published: 28 November 2019


Abstract: Electromyography (EMG) devices are well-suited for measuring the behaviour of muscles
during an exercise or a task, and are widely used in many different research areas. Their disadvantage
is that commercial systems are expensive. We designed a low-cost EMG system with enough accuracy
and reliability to be used in a wide range of possible ways. The present article focuses on the
validation of the low-cost system we designed, which is compared with a commercially available,
accurate device. The evaluation was done by means of a set of experiments, in which volunteers
performed isometric and dynamic exercises while EMG signals from the rectus femoris muscle were
registered by both the proposed low-cost system and a commercial system simultaneously. Analysis
and assessment of three indicators to estimate the similarity between both signals were developed.
These indicated a very good result, with spearman’s correlation averaging above 0.60, the energy ratio
close to the 80% and the linear correlation coefficient approximating 100%. The agreement between
both systems (custom and commercial) is excellent, although there are also some limitations, such as
the delay of the signal (<1 s) and noise due to the hardware and assembly in the proposed system.

Keywords: low-cost sensors; electromyography; validation

1. Introduction
Electromyography (EMG) provides information related to muscle activity [1–3]. For that reason,
EMG devices are used in many research fields, such as biomedical, ergonomics, physiotherapy or
sports performance applications, where it is very important to assess the behaviour of the muscles
throughout the task [4] based on the changes in the electrical signal [5–10]. Moreover, this kind of
technology can be used to improve other studies [11].
One of the main problems of existing EMG technology is the high cost of commercial devices.
Some price examples of different commercial devices are shown in Table 1.
Although EMG signal acquisition can be done in different ways, nowadays one of the most used
methods is by means of superficial electromyography (sEMG), because in comparison with other
methods, e.g., needles, it is one of the less invasive methods. Some researchers [12] consider that
sEMG is as valid as other methods, taking into account that the acquired signal must be denoised.

Sensors 2019, 19, 5214; doi:10.3390/s19235214 www.mdpi.com/journal/sensors

Sensors 2019, 19, 5214 2 of 16

Andrade et al. [12] offered various alternatives, explaining the positive and negative aspects to develop
the filtering of the signal in their research, from a LPF (low-pass filter), wavelets, noise cancelation of
the 50/60 Hz band or empirical mode decomposition (EMD).

Table 1. Price comparison between commercial systems and the low-cost electromyography
(EMG) system.

Equipment Prize (€)

Delsys Trigno (around) 20,000
Commercial Device
Cometa (around) 15,000
Bitalino Up to 150
Low-Cost Sensors
Myoware EMG + Arduino Mega 100

Several researchers have tried to build or implement low-cost EMG systems. Supuk et al. [13]
designed, developed and evaluated a low-cost EMG system, which can be used to measure the muscle
activity during human motion, focused on the design of a cascade bio-amplifier that reduces the
noise of the signal as a first step. Later they used different approaches to denoise the output signal.
Validation was only centred on the gait analysis, where researchers measured the activity of six
main muscles. Another example is the manuscript of Sophia Heywood et al. [14], where the authors
compared the signal from a low-cost EMG system and a wire-commercial device. The evaluation
consisted of the acquisition of the Vastus Lateralis signal while volunteers executed different exercises.
After the authors denoised the signal using a few filters, they carried out an evaluation by means of the
Teager–Kaiser energy operator (TKEO) and the maximal voluntary contraction (MVC) of the muscle,
with good results. A third example is the work undertaken by Cheney et al. [15] where the authors
developed the ability to adjust diverse variables, such as the gain, attenuation or offset. For that reason,
they developed a 2-channel EMG board to gather a signal that later was filtered by low-pass filter.
Unfortunately, the authors did not explain how the validation test was done.
Examples of applications of this technology cover a wide variety of fields, from the definition
of the movement of a hand [16], up to the recognition of gestures inside a vehicle [17], in the field of
rehabilitation [18–20], or even activity quantification during a deep brain stimulation intervention [21].
Up to the present, to our knowledge, there are no studies that compare the signal from a low-cost
system and a wireless custom device, and identify limitations in the operation of the low-cost system.
Thus, one of the first objectives of the present study is the comparison of different sEMG signals gathered
from two systems: a low-cost system designed to be a user-friendly system [4] and a commercial system.
Results of that analysis are intended to validate the low-cost sEMG system in order to examine whether
the low-cost system was appropriate or if it needs improvements. For this purpose, an experiment was
developed based on previous research by other authors who have used low-cost systems [13,14,16] or
have tried to get a reliable signal from them [12,15].
In order to obtain both signals an experiment was designed in which diverse subjects executed
isometric and dynamic exercises while EMG signals from the rectus femoris muscle (RF) were
simultaneously gathered by a wireless commercial device (Delsys Trigno Wireless EMG System)
and the designed low-cost system. Both signals were then analysed, and aspects of the comparison,
including system limitations were explained.
Over the course of the study, the following hypotheses have been investigated:

Hypothesis 1. The low-cost system can provide a sEMG output signal that makes it possible to discern the exercise.

Hypothesis 2. Based on previous studies [22] and related to muscle behaviour, the more difficult to keep the
position of the exercise, the higher output voltage of the low-cost system and the commercial device. According to
some authors (e.g., Moritani and Mauro [23]) this behaviour and progressive increase of the signal is connected
with the recruitment of new motor units and their spike amplitudes.
Sensors 2019, 19, 5214 3 of 16

Hypothesis 3. The signal from the low-cost system is reliable enough to be used in simple exercises.

2. Materials and Methods

This section introduces the experimental approach carried out to test the hypotheses explained
in the present work. Section 2.1 explains the steps by which the experiment was split. Section 2.2
describes the equipment used and how it was designed. Finally, Section 2.3 deals with the analysis
method, where the filtration tools and assessment of the indicators are explained.

2.1. Experiment Scenarios

The experimental work focused on the acquisition of sEMG signals from the RF muscle of various
subjects by means of two different devices. The first was a commercial device with its features indicated
in Table 2, and the second device was a system built with various low-cost components, designed and
deployed with a custom software interface using Matlab and Simulink [24].
The main purpose was to compare the sEMG signals of the low-cost system and the commercial
device in order to be able to evaluate the feasibility of the custom system and provide a certain level
of reliability.
More than 100 different cases were compared using 5 experimental subjects (3 males and 2 females)
who were selected and participated in the experiment. The average participant characteristics were:
age, 26 ± 2.9 years; weight, 61.8 ± 13.1 kg; height: 170 ± 6.2 cm.
Figure 1 shows the protocol and steps that each of the volunteers followed in order to complete
the experiment.
Step 1: the experiment and possible risks were explained to the volunteers. Each participant
had to sign an informed consent form, where details about the experiment were described to start
the exercises. Each volunteer had the chance to leave the experiment whenever he or she wanted.
In addition, although the informed consent form contained all the information, instructions to perform
the exercise were described to each of the volunteers.
Step 2: the subject completed an electronic questionnaire. This electronic questionnaire was
designed to collect certain data such as age, height or weight.
Step 3: in order to properly place the electrode sensors on the skin, a palpation test [25] was
carried out. Once the RF was located, and before attaching the electrode sensors, the area was carefully
cleaned, first by shaving the area with a disposable razor blade and then by cleaning it with alcohol
and sterile muslin.
Step 4: execution of the exercises, as explained in Figure 1.
All the exercises carried out in the step 4 were explained, as follows. All these exercises were
selected based on expert opinion, and took into account that the RF must be actively working during
each exercise [26,27].

• Squat: An exercise that has demonstrated reliability in other studies and activates the muscle with
an easy movement and minimal equipment [28].
• Lunge: A good exercise to strengthen the quadriceps. It is a also a basic and simple movement to
do for beginners [29].
• Knee Extension: this exercise is able to be done using only the muscles of the quadriceps, making it
possible to isolate the activation effect on associated muscles and focus on the selected muscle [30].
• Jump: This exercise combines the characteristics of the squat and sums movement up, therefore
the analysis is more exhausting. Additionally, it is frequently used to measure the EMG signal in
other studies [14,31].
Sensors 2019, 19, 5214 4 of 16
Sensors 2019, 19, x FOR PEER REVIEW 4 of 17

START

STEP 1 Signing of the informed consent

Experiment explication

STEP 2 Main characteristics

Online questionnaire of the volunteers

STEP 3 Sensors placement on

Palpation test the Rectus Femoris

STEP 4
Exercise Performance EMG data

FINISH

Figure1.1.Diagram
Figure ofthe
Diagram of theexperiment.
experiment.
Sensors 2019, 19, x FOR PEER REVIEW 5 of 17
Sensors 2019, 19, x FOR PEER REVIEW 5 of 17
Table All 2. Description
the Sensors 2019, 19,
exercises of the PEER
x FOR
carried isometric
out REVIEW contraction exercises (top), and description of the dynamic
in the step 4 were explained, as follows. All these exercises 5 of 17
were
Sensors 2019, 2019,
19, x FOR
19, x PEER REVIEW 5 of 17
Sensors 2019,Sensors
19, x FOR
exercises 2019, 19, Sensors
PEER
(bottom). x REVIEW
Table
FOR PEER 2. Description
Images REVIEW FOR PEER ofbyREVIEW
the isometric contraction exercises (top), and description 5 of 17 of the
5 of dynamic
17 5 of 17
Sensors selected
Sensors 2019, 19,
19, xx FOR
Sensors
2019, based
2019,
FOR PEER
Sensors
PEER on
Table
19, Sensorsexpert
x REVIEW
Sensors2019,
FOR
REVIEW 19,19,
PEER
2019, FOR xcreated
opinion,
2.x Description
2019, 19,
REVIEW
x FOR FOR
PEER
PEER PEER and
of
REVIEW theMiguel
REVIEW
REVIEW took
isometric Gómez
into account
contractionPalacios
that the
exercises[32].
RF must
(top), andbedescription
actively
55 of
of 17
17 working
of the 17 during
5 ofdynamic 5 17
5 5ofof17 of 17
Sensors 2019, 19, x Sensors exercises
FORTable2019, REVIEW
PEER 19, (bottom).
2. Description
x FOR PEER Images
of thecreated
REVIEW isometricby Miguel Gómez
contraction exercises [32]. and description 5ofofthe
Palacios(top), 17 dynamic5 of 17
Sensorsexercise
each 2019,
Sensors
Sensors Table
19,2019, exercises
x [26,27].
FOR
Sensors
19,
2019, PEER
x
19,2.x Description
2019,
FOR
Table (bottom).
REVIEW
19,
PEER
FOR 2. xDescription
PEERFOR
REVIEW of the
Images
PEER
REVIEW REVIEW
of isometric
created
the by contraction
Miguel
isometric Gómezexercises
contraction Palacios (top),
exercises [32] . and description
(top), and of the
description of dynamic
5 of 17
the5 of
5 17
of
dynamic17 5 of 17
Table 2. Description of the isometric
Table 2. Description
exercises of the contraction
(bottom). isometric
Images created exercises
contraction (top),
exercises
by Miguel
Isometric and (top),
Gómez
Contraction description of .the dynamic
and description
Palacios [32] of the dynamic
Table
Table 2.
2. Description
Table 2. Images
Description of Table
of
Table the
exercises
Description
Tablethe2.2. 2.ofDescription
isometric
Description
(bottom).
isometric
exercises contraction
theMiguel
Description ofof
Images
isometricthe
contraction
(bottom).
of
the
Images
the isometric
exercises
isometric
created by Miguel
contraction
isometric
exercises
created
contraction
(top),
contraction
Isometric Gómez
exercises
contraction
by(top), . and
Miguel
exercises
andContraction
description
exercises and(top),
Palacios
(top),
description
exercises
Gómez [32] of.(top),
of the
and
description
(top), the
and andof description
dynamic
description
dynamic
. of the dynamic
description of dynamic
ofofthe
the the dynamic
dynamic
exercises (bottom).
Tableexercises Table
(bottom).
2. Description created
2.Images
of the by
Descriptioncreatedof
isometric Gómez
bythe Miguel
contraction Palacios
isometric
Gómez [32]
contraction
Palacios
exercises
Isometric [32]
(top), . Palacios
exercises
and
Contraction (top),
description [32]
and description
the dynamicof the dynamic
•
exercises
exercises Squat:
Table 2.◦An
(bottom).
Table
exercises
(bottom).
Exerciseexercises
exercise
Description
Images
2.
Images
* N(bottom).
Duration
exercises
Table
exercises createdof
Description
(bottom).
Table 2.
exercises
rep exercises
2.that
Images the
Description
created(bottom).
by ofhas
(bottom).
Description
(bottom). by Miguel
Images
the
created
of
Miguel demonstrated
isometricImages
ofcontraction
Gómez
isometric
the
Imagesby thecreated
created
Miguel
isometric
Duration
Gómezcreated isometric
Palacios
by byreliability
Miguel Miguel
[32]
Isometric
contraction
Gómez
contraction
Palacios
by Miguel
[32] Gómez
Palacios in
.. exercises
exercises (top),
Gómez
contraction andother
Palacios
exercises
Palacios
[32] .
Contraction
exercises
Gómez
Instructions (top), [32]
and
(top),
Palacios studies
description
.(top),
[32] [32]
and
. ofand
.description
and
descriptionthe activates
dynamic
description
of the
of of the
dynamic
the the muscle
dynamic dynamic
Description
Exercise Images
(Seconds) (bottom).
created
*Nº rep Images
byDuration
Miguel createdGómez Isometric
by Miguel
Palacios
IsometricContraction
[32]. Contraction
Gómez Palacios [32].
Instructions Description
exercises
with an (bottom).
exercises
exercises
Exercise Images
exercises
easy movement
(bottom).
(bottom).
*Nº Isometric
created
(bottom).
Images
Images
rep by
Images
andcreated
minimal Contraction
Miguel
Isometric
created
Duration
(Seconds) byGómez
bycreated
equipment
Miguel Palacios
byGómez
Miguel
Contraction
Miguel Gómez [32]
Gómez
[28]. . Instructions
Palacios
Palacios[32] . . [32].
Palacios
[32] Description
Exercise *Nº rep Isometric
Isometric Duration
(Seconds) Contraction
Isometric
Contraction Isometric Isometric
Contraction
Isometric Contraction Contraction
Contraction Instructions Description
• Lunge: Exercise
A good Duration *Nº
exercise rep Duration
Isometric Isometric
Contraction Contraction Instructions basic and simpleDescription
Exercise
Squat Exercise*Nº rep 3 Squat Exercise
*Nº Duration
rep
Duration 30
Duration
Duration*Nºtorep strengthen
(Seconds)
Isometric
Duration
(Seconds) Duration
Duration
the
Isometric
Isometric
Contraction
quadriceps.
Contraction
Instructions Isometric
Contraction
Contraction
Instructions
of 90
It
◦ Contraction
knee
is Instructions
a also a Description
flexion Description
movement
Description
(Seconds)
Exercise 3
*Nº 30
(Seconds) Contraction of 90° knee flexion
Exercise
Exercise to*Nº
Exercise do rep
Exercise
*Nº Exercise
for
rep *Nº
*Nº
beginners
Exercise
Squat(Seconds)
Exercise
rep
*Nº
rep Duration
*Nº
Duration
*Nº
rep
(Seconds)
[29]. 3 reprep
rep
Duration
(Seconds) 30
Duration
(Seconds)
Instructions
Instructions Instructions
Contraction
Instructions
Instructions
Instructions
Instructions
of
Instructions 90° knee Description
Description
flexion Description
Description
Description
Description
Description
Description
Squat(Seconds) 3 Duration
(Seconds) Duration
(Seconds)
30 Contraction of 90° knee flexion
Squat• Exercise
Knee
Squat
Exercise *Nº
Exercise
Squat
3 Extension:
rep
Exercise
3 30this
Squat
(Seconds)
*Nº*Nº rep 3*Nº3rep is
rep
exercise
(Seconds) 30(Seconds)
(Seconds)
(Seconds)
30
able
(Seconds)
Contraction to Contraction
30 be done Instructions
90°Contraction
ofContraction using
knee of
Instructions
only
Contraction
flexion
90° ofknee
the ofInstructions
Instructions
the 90° ofknee
muscles
forward
flexion 90° knee flexion
leg oftoflexion
the Description
kneeDescription
90°quadriceps,
Description
Description
making
Lunge 3 30 Contraction of the forward leg to 90° knee
Squat
Squat
Lunge Squat
Squat 33 3 Squat
it possible LungetoSquat
Squat
Squat
3
30
3isolate
3030 the 333 30
30 3
3
activation Contraction
30
Contraction
30effect
Contraction
30
30
30 Contraction
Contraction onof
of 90°
ofContraction
90° knee
Contraction
knee
associated
the forward
of
ofContraction
Contraction
Contraction
90°
flexion
90°
flexion
leg
knee
knee
ofmuscles
the
to of
90 of◦90° ofand
flexion
90°
forward
flexion
of
flexion knee
90°
90°
knee
knee knee
focus
leg
flexion
knee toflexion
flexion
flexion
flexion on
90° the
knee selected muscle
Squat Squat
[30].
Lunge
Squat
Lunge 3Squat 3 3 3330 Contraction 3 30 303030 Contraction theContraction
ofContraction
30 forward
Contraction
of the ofleg
Contraction 90°of
Contraction
forward toof the
knee
90°
90° forward
offlexion
flexion
the
knee
legknee
90°
flexion of 90°
forward
knee
to
leg
flexion
90° kneeto
flexion
knee leg 90° toknee
flexion 90° knee
Lunge Lunge3 Knee Lunge
3 30 3
30Contraction 30
of
of the
Contraction Seat
forward
Contraction
of ontheaaleg
Contraction chair,
forward to
of 180°
90°
ofthe of the
knee
forward
flexion
leg knee
to forward
90°extension
leg legto
knee leg
to90° toand 90°
knee leg
knee
Lunge 33 Lunge Knee Lunge
3 30 3 33 30Contraction
3 3030
Contraction 30 theflexion
forward
ofContraction
Seat on
Contraction
the leg
chair, toof 90°
the
180°
the knee
forward
flexion
knee
forward extension
leg to 90°
90°and knee
kneeleg
Lunge • Lunge
Knee Lunge
Jump: This Lunge
Knee
extension
Lunge
3 30 combines
exercise 30 33 the 30characteristics
30
Contraction
30 Seat
flexion
Contraction
◦ of
Contraction
flexion onforward
the offlexion
aof
forward
Contractionflexionthe
chair,in the
the
of
leg
squat
180°
horizontal
leg
of
forward
to
the
flexion
forward
90°
and
knee
to
flexion 90° knee
forward
flexion
leg sums
extension
position
knee
to
leg 90°
to legmovement
knee
90° toand
knee 90° leg
knee up, therefore
Knee Lunge Lunge Knee
extension
3Lunge 3Knee30
Lunge 3 3 3330Seat 3 30
Seat on
on30 30
aachair,
chair, 30180 180° Seat
knee knee on a on
extension
flexion
Seat chair,
extension in and
a chair, 180°
horizontal knee
legflexion
and 180°inleg extension
position
horizontal
knee and and
position
extension leg leg
extension Knee
the extension
analysis is more exhausting. 3 on aa Seat 30 Seat on
Additionally,
30on180° a chair, Dynamic 180°
itflexion
is knee
in
frequently extension
horizontal
flexion
flexion used
flexion and
position tolegmeasure the EMG signal
Knee
Knee Knee Knee3 extension Knee
Knee 3 30
Knee Seat
30Seat chair, aSeat knee
chair, Seat
onon a on
extension
180° chair,
in aexercise
knee chair,
180°
horizontal and 180°
extensionkneeleg knee
extension
position
and extension
leg and andandleglegleg
extension extension 33 studies extension
Knee 3 30
30 3 3 30 3 on Seat30chair,
30 in30
on a 180°
horizontal
chair, Seat
knee
Seat
in Dynamic
180° on extension
positionaa chair,
knee
horizontal
chair, in
exercise 180°
and
horizontal
180°
extension
position
knee
leg
kneeand extension
position
extension
leg and leg
extension
extension in other
Knee Knee
extension Exercise
extension
Knee
extension Knee *Nº
3extension [14,31]. 303rep rep **reps
Seat30inin30on aSeat
horizontal
Seat chair,
Dynamic
horizontal on Dynamic
inon Instructions
180°
aSeatposition
chair, on
a chair,
horizontal
exercise
position knee
in 180°exercise
extension
ainhorizontal
chair,
in
180° knee
position
horizontal 180°
horizontal
knee andposition
knee
position
extension
extension
position leg
extension
andand Description
and leg
legDescription
leg
extension Exercise
extension3 3 *Nº 3 30 30**reps
3Dynamic 30 Dynamic exercise inin Dynamic Instructions
horizontalDynamic exercise
in horizontal
position exercise position
extension extension Exercise
extension extension *Nº repDynamic **reps 3 slow exercise
Instructions
horizontal
in
speed horizontal
in down position
inexercise
horizontal horizontal
and position
position
slow position Description
Exercise The *N ◦Exercise
rep ** *Nº
reps rep **reps
Instructions exercise
Dynamic Dynamic Dynamic
Instructions
exercise exercise Description Description
Exercise Exercise *Nºexercises
rep Squat Exercise
*Nº **repswere **reps
rep
split 3 into
*Nº Dynamic
rep two
1/10**reps
Instructions
Dynamic categories:
exercise
3 slow exercise
Instructions
Dynamic
speed
Dynamic isometric
Instructions
down exercise
exercise and
and slowdynamic
Description exercises. The isometric
Description Description
Exercise *Nº rep Squat
Exercise
**reps 3*Nºreprep 1/10 Dynamic
**reps
Instructions 3Dynamic
slow speed
exercise speed
Dynamic down
Instructions andDescription
upexercise slow Description Description
Exercise Exercise
exercises
Exercise *NºwererepExercise
*Nº
*Nº
Exercise
Squat
Exercise
rep
rep
**reps
used when *Nº **reps
*Nº
*Nº
**reps 3
rep
focusing
3 rep
slow
**reps
Instructions
speed
**reps
on
1/10
**reps theInstructions
3recording
Instructions
down slow
and 3
Dynamic Instructions
speed
slow
exercise
Instructions
of
speed
speed
exercise
down
Instructions theup down andDescription
MVC, slow
and and slowthe dynamic
Description
Description
Description
exercises
Description focused
Squat Exercise Exercise
3 Squat
*Nº
Exercise rep
Exercise
Squat*Nº
1/10 *Nº **reps
rep 3*Nº**reps
rep 33slow
rep 31/10
**reps slow
speed**reps
1/10 speed
down and
Instructions down
Instructions and
speed
Instructions slow
Instructions up Description Description
Description Description
Squat Squat
on the ability 3 of the3 1/10
system 33 slow
1/10 to
slow speed 3
accurately
speed slow down
down 3
speed 3and
slow
register
and 3
slowdownslow
speed
slow speed
the speed
speed
and downdown
behaviour up
slow down and and and
slow
of slow
the slowwave.
Squat
Squat Squat Squat3 3 Squat
Lunge
Squat Squat
1/10
3
1/10 3 3
3 3 slow
1/10 speed
slow
1/10
1/10 speed
1/10
speed up up
3down
slowup
3speed
slow speed
speed
and slow speed
down and
down upand slow slow
In Table Squat
Lunge
2,3 both 1/10of 33 exercises 3 slow
speed
1/10
1/10
speed speed
3are up
slow down
3 slow 3 speed
speed
3 3explained.
up slow
speed slowup and
speed
speed
downspeed
down slow
speed
down up
and down
up and up
and
slow and
slow slow slow
Squat Squat Lunge
Squat 3Squatsets 3 31/10 33slow3 31/10 1/10
speed 1/10 1/10
down speed slow
and 3up speed
slow
Further
speed
speed
speed down up
up
down and detailsslowslow
and
of each exercise are included in
Lunge 3 Lunge Lunge
1/10 3 31/10
slow 1/10 speed
speed downup
speed speedand
speed
up speed
up slow up up
the same
Lungeline. Knee 3 33 slow
1/10
slow speed
3 slowspeed3 slow
speed
speed
down
down 3
speed
down 3and
slow
and
up 3speed
3
slowdown
and slow
speed
slow speed speed
speed
and downdown
speed up
slow down and
upand and
slowslow slow
Lunge
Lunge Lunge
Lunge 3 3 Lunge Knee Lunge
1/10
3 1/10 31/10 3 33 slow 1/10
1/10 1/10
speed 3down
slowup
slow speed
speed
and slow down and
down and slow slow
Lunge 3 extension Lunge
3 31/10 1/10
Knee
Lunge 33 3slow
slow
speed
1/10
1/10
speed 3speed
speed up
slow up down
3 slow
3 3slow
up 3 speed
speed
speed slow and
speed
speed
up downspeed
down slow
speed
down up
and down
up and up
and
slow and
slow slow slow
Knee Lunge Lunge
Knee3 extension
Knee
extension
Lunge Lunge
Knee 3 31/10 3 3
3 31/10
slow 1/10
speed3
1/10
1/10
slow
1/10
down speed
speed
speed
slow
and 3up
downup
speed
slow
speed speed
speed
speed
speed
and
speed
up
down
speed
up slow
up
up
down
up
and
up
slowslow
and
Knee
Knee Knee Knee extension Knee
Knee 1/10
Knee
3 331/10
slow
slow 3 speed 3speed
slow 1/10
down 3 3and
speed slow 33down
slow
slow
speed down
speed
speed
and down up
slowtodown 90°
and knee
and
slow slow
extension
Knee extension 33 extension
Knee 1/10
3
1/10 31/10 3 33slow 3speed
slow 1/10
1/10
speed
down
1/10
speed up
down
andslow
3down3and
slow
speed33slow
flexion
slow
slow speed
down
speed
and
up down
and slow down
speed
down
jump
to 90°
todown
up
90°
as
and
and knee
knee
high
slow
slow
extension
extension Knee Knee
extension extension
3Knee
extension Knee
3extension 1/101/10 3 3 slow
speed
1/10
speed 3 speed
up
slow
3
up slow
speed
3 down
speed
slow slow
speedup down speed
speed
down down
speed slow
speed up
and
to up and
90° up
slow andasslow
slow
knee
extension 3 3 31/10 31/10 slow 1/10 flexion and jump upas to high as
extension
extension extension
extension
extensionJump
extension 33 slow 1/10
down 3 speed
2/10 slow speed
to up
90°
down
speedflexionupup
knee
possible. 3toslow
speed 90°
and
speed
speed
Once
up
down
knee
jump
speed
up landed asup
90°slow
high knee
as
Jump 333 slow
slowand down
down 3jump
2/10 slow to
to 90° down
90° 3 3knee
flexion slow
possible.
knee 3toand
slow slow
down
90° Once
down down
jump
knee tojump
to90°
landed to
as90° high90°
knee slow
knee knee
as
Jump flexion
3 3 slow
flexion
2/10 down as
and high
3
to
jumpflexion
slow
possible.
down 90°
◦ kneeto 90° as down
knee
as and
high
Once knee to
as
landed 90°
flexion asknee high
slow and as
JumpJump flexion
3flexion 3 slow and
and 3
down
2/10 slow
jump
flexion
jump todown
3 and90
as high
flexion
possible.
slow
as3flexion
down slow
jump
high to3
flexion
down to90°
slow
as
and
down
as Once
as
90°
and knee
to and
highdown
jump 90°
to
knee
jump jump
landed
90° as
knee
as tohigh
knee
flexion
as 90°
asslow
high knee
high
as
andas as
Jump Jump3 2/10
3 possible.
2/10 3 flexion
flexion Once
possible.
and
2/10
and
jumplandedjumpflexion
Once
down
asreturn
high
possible.
slow
as and
high
landed
to to90° stand Once
jump
as
slow
knee up landed
as
flexionhigh
position and slow
as
Jump
Jump Jump Jump33 Jump Jump Jump
2/10
3 3possible.
2/10 3 flexion Once
2/10
possible.2/10 and
landed
flexion down
flexion jump
possible.
Once and
return flexion
slow
to
andas
possible.90°
jump
landed high
toOnce and
knee
jump
stand Once
as
slow asjump
landed
as landed
flexion
high
up high asasslow
position ashigh
and slow as
Jump Jump Jump
3 2/102/102/10
Jump
3Jump3Jump
downpossible.
33 as to 90°Once
*possible.
Nº down
possible. 2/10
rep:knee
2/10 landed
to
OnceOnce
number 90°
possible.
flexion slow
down
possible.
landed
knee
landed
return
of and to to Once
Once
slow
flexion
repetitions;
90°
stand and
landed
knee
landed
up
** position
reps:
slow
flexion
slow
repetitions
and
per second.
3 32/10
down
downreturn
32/10
to
to
* Nº
slow
possible.
90°
2/10
down
90°rep:
todown
knee
knee
stand
2/10 to
number
toup
Once
flexion
down
possible. return
possible.
90°
flexion
down
90knee◦ down
knee landed
possible.
to
ofreturn
knee
position
and
Once
and
to 90°to
stand
Once
flexion
90°
repetitions; slow
90°
knee
landed
toand
Once
landed
knee
stand
knee
up
and landed
flexion
flexion
** reps: flexion
position
slow slow and
up repetitionsand
position
slowand per second.
down
* Nº rep: toto
return 90°
number to down
standof to
up90°
flexion position
repetitions; knee ** flexion
reps: and
repetitions per second.
return down
Nºto stand down 90°
up
down kneedown
position
return
to return
90°
to flexion
toknee
90° to
stand
knee 90° and
toflexion
stand knee
**up
flexion up flexion
position
and and and
position per second.
2.2.* Nº rep: number
Equipment * Nº rep:
return *flexion
ofnumber * return
torep:
repetitions;
return stand
Nºof number
andrep:
to
to
up
**return
number
reps:
stand
repetitions;
stand
ofto
position
return up
repetitions;
of
repetitions
return
up **
to position
stand
repetitions;
to
position
reps: stand per up
repetitions
reps:
up repetitions
position
**position
second. reps:
per repetitions
second. per second.
2.2.* Equipment
* Nº rep: number
Nº rep: number * Nº rep: of return
stand*
repetitions;
*
ofnumberNº Nº
rep:
repetitions;
* Nº rep: up rep:
number to
** stand
position
returnnumber
return
reps:
of repetitions;
number
** reps: ofto up
return
stand
toof
repetitions
of position
repetitions;
** reps:
repetitions
repetitions; to
up
repetitions;
stand stand
per position
up
**
repetitions
per **up
position
second.
reps: reps:position
repetitions
reps: per
**second. repetitions per
second.per second.
repetitions per
second. second.
2.2. Equipment * Nº rep: number * Nº ofrep: number **
repetitions; ofreps:
repetitions;
repetitions ** reps: per repetitions
second. per second.
2.2. Equipment
The * Nº
testbed rep:
◦
* N rep: *used
Nºnumber
rep: was * of
Nº
number repetitions;
builtrep: using
ofnumber
repetitions;**
two reps:
of repetitions
different
repetitions;
** reps: EMG** per
reps:
repetitions second.
systems.
repetitions
per The
second. first
per system comprised a set
second.
usednumber of repetitions; ** reps: repetitions per second.
* Nº rep: number of repetitions; ** reps: repetitions per second.
2.2. Equipment2.2.
2.2. Equipment The Equipment
testbed was built using two different EMG systems. The first system comprised a set
2.2.
2.2. Equipment
2.2. Equipment
Equipment 2.2.
of
2.2. 2.2.
Equipment
The
low-cost Equipment
Equipment testbed
elements, usedand wasthe built second usingsystem two different was a commercial EMG systems. solution Thesupported
first system bycomprised
the Mechanical a set
2.2. Equipment 2.2.
of The testbed
Equipment
low-cost The elements,
testbed used was
and
used built
the
was secondusing
built two
system
using different
two was EMG
a commercial
different firstEMG systems. solutionThe The first
supportedsystem bycomprised
thecomprised a seta set
Mechanical
The testbed
2.2. 2.2. used
Equipment
The2.2.testbed
Equipment
of was
2.2.
Equipment
low-cost
Engineering built
Equipment
used using
was
elements,
Departmentbuilt two and different
using the
of two
thesecond EMG
different
University systemsystems.EMG
ofwas The
systems.
a commercial
Manchester. The systems.
system first comprised
solution system first
a setsystem
comprised
supported by a set
the Mechanical
The testbed
The testbed used
of
Theusedtestbed was
The
low-cost
Engineering
wasThe The
built
testbed
used
built testbed
testbed using
elements,
was used
Department
using built
used used
two
two was
and using
was was
different
thebuilt
different second
two
of built
the
built built EMG
using using
different
University
using
EMG two
system two two
systems.
different
systems. was
EMG
of different
The
aThe
Manchester.
different EMG first
commercial
systems. first
EMG EMG system
systems.
The
system systems.
comprised
solution
systems. first The
system
comprised
The The
first
supported
first first
a
systemsetsystem
set bya the
comprised
asystem
system comprised
comprised
Mechanical
a set
comprised a seta set
of low-cost elements,
ofThe testbed
low-cost of
and
elements,
Engineering low-cost
The
used
Boththe second
testbed
was and elements,
builtthe
Department
devices system
used using
second
were was and
was
two
of
placed the
system
the second
adifferent
commercial
using
was
University
on the system
atwo
EMG
same solution
different
commercial
of was
systems.
Manchester.
region asupported
of commercial
EMG The
solution
the systems.
first
muscle by solution
the
toThe
system
supported Mechanical
get firstsupported
comprised
by the
acomprised Mechanical
synchronized by
set the
comprised
signal aa set
from
set
Mechanical
of
of low-cost
low-cost of The
elements,
low-cost
elements,testbed
ofThe andof
low-cost
Engineering
The
ofand used
the
testbed
elements,
Both
low-cost The
low-cost
testbed
the was
testbed
second
elements,
used
and
devices
second used
elements,built
elements,
Department
the system
was using
used
and
was
second
were
system built
and was
and
was
the
of
built
placed
was two
the the
using
system
the a
second different
builtsecond
commercial
University
using two
on the
asecond
second was
commercial using
system
two EMG
different
a
system two
system
of solution
was systems.
different
was
a
Manchester.
different
commercial
samesolution region
was EMG EMGofa
commercial The
EMG
commercial
supported
systems.
solutionsystems.
the muscle first
systems.
by
solution
The system
solution
the
The
supported
by to first
theget The
Mechanical
supported
system
first by first
supported
system
the system
by
comprised a
the set
by
comprised
Mechanical
a synchronized comprised
the a
thesignal set
a from set
Mechanical
Mechanical set a
Engineering
ofof Department
low-cost
Engineering of Engineering
thelow-cost
elements, of
Department
Both
same the
and elements,
the
devices
muscle ofDepartment
University
secondthe
were
during and of
system
University
placed of
Manchester.
the
exercise. the
was
on ofTheUniversity
thea system
commercial
Manchester.
same
muscles was
region aasolution
ofselected commercial
supported
Manchester.
commercial
of the was supported
muscle
solution
thesolution
RF todue
Mechanical
by
get
supported
supported
the
a the
to highby
Mechanical
synchronized
its
by the
number
Mechanical
Mechanical
signal
of from
motor
Engineering
Engineering low-cost
Department
of
Engineering elements,
low-cost
of
Department of
Engineering
low-cost
the Both
same
Engineeringlow-cost
Engineering
of and
the
elements,
Department
of elements,
muscle
the the
elements,
University
Department
devices ofand
University
Department second
Department
were and
the
during the ofof
placedsystem
and
second
the
University
of the
of
Manchester.
the
second
exercise.
ofManchester.
the the was
second
University
on the
system
of system
The amuscles
University
same commercial
system
was
Manchester.
University of
wasregion
a
of of was
Manchester.
commercial
a commercial
selected
Manchester. solution
Manchester.
of athecommercial
wasmuscle supported
solution
solution
the to
RF solution
get
supported by
a
supported
due to its Mechanical
supported
synchronized
by
highthe
by by the
signal
Mechanical
the
number Mechanical
Mechanical
of from
motor
Both devices
Engineering Bothwere devices
the
units placed
Engineering
Department
same Both
were
[33–36], on
of
muscle devices
the
Department
the
placed
somethingduringsame were
University
on ofplaced
region
the the
exercise.
that, ofof
same ofThe
according onthethe
University
Manchester.
region muscle
muscles
to same
of ofthe
the to region
Manchester. get a of
muscle
selected
instructions tothe
synchronized
was get
of themuscle
the a RF to
synchronized
due
manufacturer get
signalto itsafrom
synchronized
signal
high
of the from of
number
low-cost signal
motor
sensor, from
BothEngineering
devices
Both muscle were
the
Engineering
devicesBoth Department
EngineeringEngineering
devices
units
were placed
Both
same placed
Both Both on
devices
muscle
Department of
devices
Department
were
[33–36], devices
on the
placed
something
the the
Department
were
during wereUniversity
same
of
same were
on the
of region
placed
exercise.
the of
placed
University
the
that,
placed
region the
onon of
University
same
according
of Manchester.
the
The University
on
the
region
the the
ofmuscle
same
muscles sameregion
Manchester.
of
to
same
muscle of of
to
Manchester.
the the Manchester.
region
selected
tomuscle get
instructions
region ofofathe
get of
a the
the the
synchronized
was tomusclethe
get
of
synchronized muscle
the
muscle aRF totodue to
signal
get
synchronized
manufacturer
get get
signal its a synchronized
from
a asynchronized
to highsignal
of
synchronized
from number
the signal
from
low-cost signal
from
of motor
signalsensor,
from from
the same Both
the same during
devices
muscle
units
makes theBoth
weresame
exercise.
during
[33–36],
better muscle
devices
placed The
exercise.
something
the were
on
measure during
muscles
the placed
Thesame
that,
of exercise.
selected
muscles
a on
region
according
cleaner the The
was same
of
selected
to
signal muscles
the
thethe RF
region
muscle
was due
instructions
with selected
the
lessofto
to RF its
get due
ofwas
high
muscle
a
interferences the toRF
number
synchronized
the to its get
high
manufacturer
and dueof to
motor
anumber its
synchronized
signal
crosstalk of high
thefrom
of number
motor
[37].low-costsignal offrom
sensor,motor
the same the Both
muscle
same devices
during
the
units
Both the
same
theBoth
muscle
makes were
[33–36],
devicesBoth
same
exercise.
muscle
devices
during
better placed
devices
muscle
The during
something
were
the were
exercise. on
placed
measure were
during
muscles the
exercise.
that,
placedThe on
of same
placed
exercise.
selected
according
the
aon
muscles region
The onsame
same
the
cleaner the
The
was
muscles
to of
region
selected
signal the
same
muscles
the RF muscle
region
due
selected
instructions
region of
was the
of selected
the theto to
of
muscle
RF its
was get
the
of
muscle
duewas
high
the
theamuscle
to tosynchronized
the
number
RF
get
toits aRF
due
manufacturer
get toadue
high toget
of its
synchronized to
motor
synchronized
number signal
ahighsynchronized
its
of thehigh from
number
of number
low-cost
signal
signal
motor signal
ofof
from motor
sensor,
from from
of motor
the
units same
the muscle
[33–36],
same
units
during
something
muscle
[33–36],the
makes
same
unitsexercise.
that,
same
during
something
muscle
[33–36],
better
Figure according
muscle The
exercise.
that,
the
2muscle
shows
during
muscles
something
during to
The the
according
measure how
exercise.
selected
that,
instructions
exercise.
muscles
of thea to the
cleaner
testing
The
according
The was
selected muscles
of
muscles
instructions
signal
the
the towith
was
equipment
RFthe
with
selected
due
manufacturer
selected
theof
less
RFto
instructions
the
less
was
interferences
its
due was
was high
manufacturerof
to
interferences
installed,
the
of
the
the number
itsitsRF
the
RF
high
the
and
due
due crosstalk
ofnumber
and the
of
to
manufacturer
low-cost to motor
its
sensor,
its
low-cost
crosstalk
low-cost
high
high [37].
of
ofsystem
motornumber
the
number
sensor,
[37]. low-cost
and of
its
motor
sensor,
motor
wire-
units
units the
[33–36],same
units
[33–36], muscle
something
the units
makes
same
the
[33–36],
something same
units the
units during
musclesame
that,
[33–36],
better[33–36],
muscle
something
Figure
that,
[33–36], exercise.
according
something
the
during
2 during
that,
shows
according something
measure
something to The
duringthe
exercise.
how
to that,
of
exercise.
according
the muscles
exercise.
that,
instructions
a
the
that, according
cleaner
The to
instructions The selected
according
muscles
the
testing
according Theof
to
signal
muscles
instructions the to was
muscles
selected
equipment
of to thethe the
selected the
manufacturer
instructions
with less
of
manufacturer
instructions the
was RF
selected
instructions was due
the of
interferences
the
manufacturerwas
of
the
RF
installed,
of
of toof
the
the due
RF
the thedue
thehigh
toRF
low-cost
manufacturer
and ofits
to number
due
manufacturer
the high
its
low-cost
manufacturer
low-cost to
sensor,
crosstalk high
low-cost
sensor, its
of of
the
number ofmotor
high
[37].
number
system
of the the number
low-cost
of
sensor,low-cost
and
low-costmotor
of motor
its of
sensor, motor
sensor,
wire-
sensor,
makesunitsbetter
makes thebetter
[33–36], measure
units makes
something
the
connection of abetter
[33–36],
measure
Figure cleaner
that,
2that,
with showsofthe
something
according
the measure
asignal
cleaner
how
computer with
that,to
the of
signal ainstructions
less
according
the cleaner
testing
and interferences
with
with to
lesssignal
equipment
the the with
theand
ofinstructions
interferences
muscle less
crosstalk
manufacturer
wasby interferences
of the
and
installed,
means [37].
of manufacturer
of
crosstalk
the
three the and
low-cost
[37].
low-cost
electrodes.crosstalk
of the
sensor,
system
On [37].
low-cost
theandrightitssensor,
wire-
side of
units
makes better makes [33–36],
the
unitsmeasure
makes
units
better something
units
makes
Figure
[33–36],
[33–36],
the of
better[33–36],
abetter
measurecleaner
the
2something
somethingshows ofthe
measureaccording
somethingmeasure
asignal
how
that, that,
cleaner with
ofthe
according to
athat,
of the
less
cleaner
testing
according
signal instructions
aaccording
cleaner
interferences
to
with signal
equipment
the
to the
less to
signal
withthe
instructions of
instructionswith
and
interferences the
instructions
less manufacturer
less
wascrosstalk
interferences
of installed,
the and of
ofinterferences the
[37].
manufacturer
the theandof
manufacturer
crosstalk theand low-cost
manufacturer
of crosstalk
crosstalk
low-cost
[37]. ofsystem
the sensor,
of
[37].
low-cost
the the
[37].
low-cost low-cost
and its
sensor, wire-
sensor, sensor,
Sensors 2019, 19, 5214 5 of 16

The exercises were split into two categories: isometric and dynamic exercises. The isometric
exercises were used when focusing on the recording of the MVC, and the dynamic exercises focused
on the ability of the system to accurately register the behaviour of the wave.
In Table 2, both sets of exercises are explained. Further details of each exercise are included in the
same line.

2.2. Equipment
The testbed used was built using two different EMG systems. The first system comprised a set of
low-cost elements, and the second system was a commercial solution supported by the Mechanical
Engineering Department of the University of Manchester.
Both devices were placed on the same region of the muscle to get a synchronized signal from
the same muscle during exercise. The muscles selected was the RF due to its high number of motor
units [33–36], something that, according to the instructions of the manufacturer of the low-cost sensor,
makes better the measure of a cleaner signal with less interferences and crosstalk [37].
Figure 2 shows how the testing equipment was installed, the low-cost system and its
wire-connection with the computer and with the muscle by means of three electrodes. On the
right side of Figure 2 is shown the commercial system. Commercial system electrodes only need to be
placed on the muscle and aligned with the muscle fibres, and its communication with the computer
Sensors 2019, 19, x FOR PEER REVIEW 6 of 17
is
by means of a wireless connection.

Low-cost device Commercial device

3
2

Figure 2. Equipment
Figure of the
2. Equipment testbed
of the used
testbed usedininthe
theexperiments.
experiments. 1.1.Arduino
Arduino Mega
Mega board.
board. 2. Myoware
2. Myoware
EMGEMG Muscle Sensor
Muscle (SEN-13723
Sensor (SEN-13723ROHS).
ROHS). 3. Delsys
DelsysTrigno
Trigno Wireless
Wireless EMGEMG System. A more detailed
System.
wired connection layout of the low system can be found in the supplementary material.
2.2.1. Low-Cost Custom System
2.2.1. Low-Cost Custom System
The low-cost system was made up of 3 components: a laptop with Matlab/Simulink [24], an
The low-cost
Arduino system
board (1 was made
in the Figure 2) andupthe of
EMG 3 components: a laptop
sensor (2 in the Figure 2). with Matlab/Simulink [24],
The Arduino board was in charge of keeping the communication
an Arduino board (1 in the Figure 2) and the EMG sensor (2 in the Figure with2).the sensors, registering
the information and then sending it to the computer, where
The Arduino board was in charge of keeping the communication with the it would be saved andsensors,
later analysed.
registering
One of the main problems of the Arduino board was its limitations (power
the information and then sending it to the computer, where it would be saved and later and memory), therefore
analysed.
the Arduino Mega was selected, because the EMG process needs a large memory to perform
One of the main problems of the Arduino board was its limitations (power and memory), therefore the
acquisition and filtration of the data. The main characteristics of the EMG chip sensor and the
Arduino Mega was selected, because the EMG process needs a large memory to perform acquisition
Arduino Mega board are shown in Table 3. This sensor needs three electrodes, one close to the middle
and filtration of the
of the muscle bodydata. Theinmain
(red line characteristics
Figure of the EMG
2), the second electrode linedchip sensor
up with and theofArduino
the direction the fibre’sMega
boardmuscle
are shown in Table 3. This sensor needs three electrodes, one close to the middle
and close to the end of it (blue line in Figure 2), and the last one placed near to a bony of area
the muscle
as
bodya(red line in Figure 2), the second electrode lined up with the direction of
reference (green line in Figure 2) to deal with the crosstalk signal from other muscles [37]. the fibre’s muscle and
close to the end of it (blue line in Figure 2), and the last one placed near to a bony area as a reference
Table 3.2)
(green line in Figure Technical
to dealspecifications of the Arduino
with the crosstalk Mega
signal fromboard andmuscles
other the low-cost EMG chip.
[37].
Arduino Mega
Microcontroller ATmega2560
Vin (V) 7–12
Vout (V) 6–20
Digital Inputs/Outputs 54
Sensors 2019, 19, 5214 6 of 16

Table 3. Technical specifications of the Arduino Mega board and the low-cost EMG chip.

Arduino Mega
Microcontroller ATmega2560
Vin (V) 7–12
Vout (V) 6–20
Digital Inputs/Outputs 54
Analogue Inputs 16
Flash Memory (Kb) 256
SRAM (Kb) 8
EEPROM (Kb) 4
Clock Speed (MHz) 16
sEMG Sensor
Supply (V) 2.9–5.7
Output modes EMG Envelope/Raw EMG
Size (cm) 2.08 × 5.23

To place correctly the sensor pads on the skin it was necessary to locate the muscle and the direction
of the muscle fibres with the help of a palpation test based on the instructions of the SENIAM (Surface
ElectroMyoGraphy for the Non-Invasive Assessment of Muscles) [38]. After that, the skin should
be cleaned of dirt and made hairless. A disposable razor was used to remove the hair, and alcohol
with a sterile gauze was applied to clean the area. Once this step was done, the three electrode pads
were attached.

2.2.2. Commercial System

The second system used was a commercial one (Figure 2). Its features are summarised in Table 4.
This equipment is used by a wide range of studies [39–42] because of its high accuracy and the
powerful software.

Table 4. Delsys Trigno Wireless EMG system. Technical information.

EMG
Dimensions (mm) 27 × 37 × 13
Mass (g) 14
EMG Signal Input Range (mV) 11/22
EMG Signal Bandwidth (Hz) 20–450/10–850
EMG Contact Dimensions (mm) 5×1
Contact Material 99.99% silver
Accelerometer
Accelerometer Bandwidth (Hz) 24–470
Accelerometer Range (g) ±2, ±4, ±8, ±16
Gyroscope
Gyroscope Bandwidth (Hz) 24–360
Gyroscope Range (dps) ±250, ±500, ±1000, ±2000
Magnetometer
Magnetometer Bandwidth (Hz) 50
Magnetometer Range (µT) ±4900

In contrast to the low-cost system, this equipment is wireless, which makes it easier to install on
the muscle and connect with the computer. Instead of 3 pads in different areas, this device only uses
one sensor that should be lined up with the direction of the muscle fibres [43].
Sensors 2019, 19, 5214 7 of 16

2.3. Analysis Method

Once all the exercises had been done and signals from both devices (commercial and low-cost)
were saved, we began with the data analysis as follows.
On one side and from a medical point of view, an EMG signal altered by noise might cause a
wrong diagnosis of the behaviour of the muscle. On the other side and from a scientific and technical
point of view, a signal affected by noise does not guarantee a good interpretation of the results, reducing
the reliability of the results obtained in research [44–46]. Therefore, due to the characteristics of the
signal, it is recommended [47] to apply some kind of filter to remove the noise of the signal before
starting the analysis.
The analysis method used was based on the combination of the methodology used by
Dufour et al. [48] and the methodology used by Heywood et al. [14], and its stages are as follows:

1. Signal processing (signal synchronization, signal trimming and signal filtering): two independent
systems were used to measure the signal, therefore it was necessary to synchronize both signals
by means of a trigger. The signal was trimmed to the time of the exercise and finally filtered to
denoise it. All the filtering processes were performed by EMGworks [49].
2. MVC computation and normalization of the dynamic signal: MVC was computed for every
exercise. It was used to normalize the signal because each system amplifies the output signal
based on its configuration (see also Isometric MVC Assessment).
3. Indicators assessment: the last stage was the assessment of the indicators in order to compare
both signals (see Validation Indicators). The analysis of the signal was made by Matlab [24] and
Microsoft Excel [50].

2.3.1. Signal Processing

This section has been split into three parts. First it is explained how the filtering was carried
out. Then it comes a short explanation about synchronizing and trimming of the signal. Finally, it is
explained the assessing of the isometric MVC.
1. Signal filtering
As explained earlier, two different EMG devices with different technical specifications were used
in the experiment. In order to denoise both signals, it was necessary to perform different signal filtering
for each one. On one hand the signal from the low-cost system was filtered in 2 steps. Each filter was
designed based on the evaluation of the fast Fourier transform and previous studies [4,14]. First a band
pass Butterworth 40–100 Hz, order 4, was used to remove the main noise from the signal (Figure 3B).
Following the low-cost system addressed some noise centred around 50 Hz, that was filtered by a stop
band Butterworth 45–55 Hz, order 4 (Figure 3C).
On the other hand, the signal from the commercial system was filtered by a band pass Butterworth
40–400 Hz, order 4. The data processing algorithm was found to be appropriate by many researchers
such as Wei et al. [51].
2. Signal synchronization and trimming
Before starting the exercises, volunteers were asked to knock the floor with their sensorized leg.
In this manner, a single precise indication would be generated to be able to overlap and synchronize
both signals, the signal from commercial equipment and the signal from the low-cost system. When
both signals were synchronized, it was necessary to determine the duration of each of the exercises
and trim both signals. This technique was semi-automatic by the use of a custom-designed script in
Matlab [24]. It is important to note that the sEMG signal was changed with a single polarity with the
purpose of ensuring an average not equal to zero (Figure 4).
 As explained earlier, two different EMG devices with different technical specifications were
used in the experiment. In order to denoise both signals, it was necessary to perform different signal
filtering for each one. On one hand the signal from the low-cost system was filtered in 2 steps. Each
filter was designed based on the evaluation of the fast Fourier transform and previous studies [4,14].
First a band pass Butterworth 40–100 Hz, order 4, was used to remove the main noise from the signal
Sensors(Figure
2019, 19,3 5214
B). Following the low-cost system addressed some noise centred around 50 Hz, that was 8 of 16
filtered by a stop band Butterworth 45–55 Hz, order 4 (Figure 3 C).

(A) RAW signal (B) 1st Filter (C) 2nd Filter

Amplitude (V)

Time (s)

FigureFigure
3. An 3. An example
example ofof thereduction
the reduction of
of signal
signalnoise
noiseusing several
using filters.
several On the
filters. Onleft
the(A)left
the(A)
RAWthe RAW
signal between second 9 and 23. In the middle (B) the signal once the first filter was applied. On the
signal between second 9 and 23. In the middle (B) the signal once the first filter was applied. On the
right (C) the complete filtered signal once the second filter was applied.
right (C)
Sensors the19,
2019, complete filtered
x FOR PEER REVIEW signal once the second filter was applied. 9 of 17
On the other hand, the signal from the commercial system was filtered by a band pass
Butterworth 40–400 Hz, order 4. The data processing algorithm was found to be appropriate by many
researchers such as Wei et al. [51].
2. Signal synchronization and trimming
Before starting the exercises, volunteers were asked to knock the floor with their sensorized leg.
In this manner, a single precise indication would be generated to be able to overlap and synchronize
both signals, the signal from commercial equipment and the signal from the low-cost system. When
both signals were synchronized, it was necessary to determine the duration of each of the exercises
and trim both signals. This technique was semi-automatic by the use of a custom-designed script in
Matlab [24]. It is important to note that the sEMG signal was changed with a single polarity with the
purpose of ensuring an average not equal to zero (Figure 4).

FigureFigure 4. Signal
4. Signal treatment
treatment procedure.On
procedure. On the
the top
top the
theabsolute
absolutevalue of the
value signal
of the afterafter
signal filtering is
filtering is
plotted (Section 2.3.1). In the middle, signal of both systems (commercial and low-cost custom) once
plotted (Section 2.3.1). In the middle, signal of both systems (commercial and low-cost custom) once
they were normalized by means of the MVC. On the bottom of the figure both signals are overlapped
they were normalized by means of the MVC. On the bottom of the figure both signals are overlapped
and trimmed according to the duration of the exercise. On the right (orange) signal from the low-cost
and trimmed according to the duration of the exercise. On the right (orange) signal from the low-cost
custom system and on the left (green), signal from the commercial system.
custom system and on the left (green), signal from the commercial system.
3. MVC assessment
Once the signals of both sets of equipment were filtered, synchronized and trimmed, the next
step was assessing the isometric MVC of each exercise and then normalizing the signal to enable
calculation of the indicators that check the viability of the low-cost system.
Sensors 2019, 19, 5214 9 of 16

3. MVC assessment
Once the signals of both sets of equipment were filtered, synchronized and trimmed, the next step
was assessing the isometric MVC of each exercise and then normalizing the signal to enable calculation
of the indicators that check the viability of the low-cost system.
Sensors 2019, 19, x FOR PEER REVIEW 10 of 17

2.3.2. Isometric MVC Assessment

2.3.2. Isometric MVC Assessment.
The maximal contraction of the quadriceps of the volunteers was achieved through three
The maximal contraction of the quadriceps of the volunteers was achieved through three
different exercises as listed in Table 2. Each one of these exercises was used to normalize the related
different exercises as listed in Table 2. Each one of these exercises was used to normalize the related
dynamic EMG signal. The output ranges of each set of equipment are quite different, making signal
dynamic EMG signal. The output ranges of each set of equipment are quite different, making signal
normalization necessary.
normalization necessary.
The diagram in Figure 5 shows the steps that have been followed for the normalization of the
The diagram in Figure 5 shows the steps that have been followed for the normalization of the
signal. On the left side of the diagram is shown the evolution of sEMG signal of the isometric exercises
signal. On the left side of the diagram is shown the evolution of sEMG signal of the isometric exercises
(knee extension, lunge and squat). Those signals were used to obtain the MVC from the waveform
(knee extension, lunge and squat). Those signals were used to obtain the MVC from the waveform of
of each exercise and from the commercial and low-cost system separately. On the right side can be
each exercise and from the commercial and low-cost system separately. On the right side can be seen
seen the treatment of the sEMG signal of the dynamic exercises, from the acquisition to the signal
the treatment of the sEMG signal of the dynamic exercises, from the acquisition to the signal
processing, where it was denoised. After that the dynamic sEMG signal was normalized according to
processing, where it was denoised. After that the dynamic sEMG signal was normalized according
the formula indicated in the same figure (Figure 5) by means of the corresponding MVC value.
to the formula indicated in the same figure (Figure 5) by means of the corresponding MVC value.

START

Isometric exercise Dynamic exercise

Signal processing
Signal processing
Isometric MVC

Knee Ext. MVC Knee Ext.

Lunge MVC Lunge

Squat
Squat MVC
Jump

Signal normalization

𝑑𝑦𝑛𝑎𝑚𝑖𝑐 𝐸𝑀𝐺 𝑠𝑖𝑔𝑛𝑎𝑙

𝑁𝑆 =
𝑖𝑠𝑜𝑚𝑒𝑡𝑟𝑖𝑐 𝑀𝑉𝐶

NS Knee Ext. FINISH

NS Lunge

NS Squat

NS Jump

Figure
Figure 5.
5. Diagram
Diagram followed
followed to
to normalize
normalize the
the sEMG
sEMG signal
signal from
from the
the dynamic
dynamic exercises.
exercises.

2.3.3. Validation Indicators

In order to achieve a successful validation of the low-cost signal, a set of validation indicators
were calculated. Each indicator was selected based on a different criterion, explained in the following
lines.
Sensors 2019, 19, 5214 10 of 16

2.3.3. Validation Indicators

In order to achieve a successful validation of the low-cost signal, a set of validation indicators were
calculated. Each indicator was selected based on a different criterion, explained in the following lines.

1. Spearman’s correlation: assesses the monotonic relationship of two variables. It can move between
−1 and 1, 1 when signals are identical and −1 when signals are fully opposed. This correlation has
been used in several studies [52,53] to compare signals from different sources with good results.
2. Energy ratio: defined as the energy (E) as in Equation (1) [54], the energy ratio moves between 0
(no similarity) to 1 (fully identical), and it compares the amount of energy step by step, testing if
the sEMG waves have the same form in both scenarios.
Z t 2
E= x(t) dt (1)

0

3. Linear correlation coefficient (LCC): the LCC has been calculated according to Equation (2); x
represents the values of the low-cost system and y the values of the commercial system. Linear
correlation is bounded between −1 and 1. When LCC > 0 it is a positive correlation. In case
LCC < 0 it is a negative correlation. If LCC = 0 there is no correlation between both variables,
and is an uncorrelated relation. One example where this correlation has been used is in the work
undertaken by Andrea et al. [31].
P
(x − x)( y − y)
LCC = q (2)
(x − x)2 ( y − y)2
P P

4. Cross-correlation coefficient (CCC): One widespread method (3) to compare EMG signals [55,56].
The CCC can stay between −1 to 1, where −1 means negative correlation and 1 positive correlation.
P P P
n( xy) − ( x)( y)
CCC = qh (3)
n ( x2 ) − ( x ) 2 n ( y2 ) − ( y ) 2
P P ih P P i

2.4. Volunteers Classification

The validation of the low-cost system was carried out using volunteers who participated in the
experiment. There were three male and two female volunteers, with an average age of 26.6 ± 2.9 years,
an average height of 170 ± 6.2 cm and an average weight of 61.8 ± 13.1 kg.
All the volunteers were informed about the experiment and agreed to use all the data for scientific
purposes. In addition, they were informed that the experiment could be stopped at any time as they
wished, with no obligation to complete it.

2.5. Validation Analysis

The dataset involved more than 100 different curves to carry out the comparative and validation
work, because each volunteer repeated the exercises a number of times. Results are shown below.
Validation analysis was split into two sections. The first section compares the isometric sEMG
output signal of the commercial and low-cost devices. After that results of the indicators (Section 2.3.3)
are explained.

3. Results
The following section presents the results obtained in this work. First it describes the sample of
the volunteers that participated in the experiment. After that it presents the results of the experiment
in two different stages.
Sensors 2019, 19, 5214 11 of 16

3.1. Isometric Exercises

As is known, muscle work activation level and the output signal are dependant according to
previous studies [36,57]. Therefore, the main aspect of this section is to show the MVC values, gathered
from the isometric exercises, in order to check if the low-cost system behaves as the commercial system
as explained before and confirms the reliability of the system.
As can be seen in the chart in Figure 6, MVC values of all the volunteers were represented. They
have been
Sensors grouped
2019, 19, by exercise
x FOR PEER REVIEW type and sorted by commercial or custom device. 12 of 17

Figure 6.
Figure MVC distribution
6. MVC distribution according
according to
to the
the exercise
exercise and equipment used.

3.2. Indicators
3.2. Indicators Assessment
Assessment
Results of
Results of the
the comparison
comparison between
between the
the commercial
commercial and and low-cost
low-cost system
system based
based onon the
the indicators
indicators
can be found below. These indicators provide information about the signal and its
can be found below. These indicators provide information about the signal and its reliabilityreliability when this
when
system
this is used
system for simple
is used exercises,
for simple as inas
exercises, the
inexperiment.
the experiment.
Table 5 shows a summary of the indicators. Parameters include
Table 5 shows a summary of the indicators. Parameters include the
the maximum
maximum (max),
(max), minimum
minimum
(min), mean
(min), mean(µ) (μ)and
andstandard
standarddeviation
deviation(σ)(σ)
for afor
better understanding
a better and toand
understanding compare better both
to compare signals.
better both
signals.
Table 5. Current validity. SC, Spearman’s correlation; ECS , energy low-cost system, ECMS; energy
commercial system; LCC, linear correlation coefficient; CCC, cross-correlation coefficient.
Table 5. Current validity. SC, Spearman’s correlation; ECS, energy low-cost system, ECMS; energy
commercial system; LCC, linear correlation coefficient; CCC, cross-correlation coefficient.
Exercise
Lunge Knee Extension
Exercise Squat Jump
max Lunge 0.96 Knee Extension
0.71 Squat
0.64 Jump
0.70
min
max 0.96 0.51 0.53
0.71 0.570.64 0.590.70
SC
µ 0.67 0.63 0.61 0.6
min 0.51 0.53 0.57 0.59
SC σ 0.2 0.08 0.03 0.1
μ 0.67 0.63 0.61 0.6
max 0.92 0.82 0.91 0.80
σ 0.2 0.08 0.03 0.1
min 0.71 0.98 0.82 0.85
ECS /ECMS max µ 0.92 0.81 0.82
0.88 0.920.91 0.80.80
min σ 0.71 0.11 0.98
0.11 0.110.82 0.60.85
ECS/ECMS
μ max 0.81 0.99 0.88
1 1 0.92 0.970.8
LCC
σ min 0.11 0.97 0.97
0.11 0.980.11 0.930.6
µ 0.99 0.99 0.99 0.95
max σ 0.99 0.01 0.01
1 0.01 1 0.020.97
max 0.87 0.83 0.72 0.71
LCC minmin 0.97 0.51 0.97
0.5 0.520.98 0.50.93
CCC
μ µ 0.99 0.68 0.64
0.99 0.620.99 0.610.95
σ σ 0.01 0.15 0.16
0.01 0.08
0.01 0.04
0.02
max 0.87 0.83 0.72 0.71
ECS /ECMS ,0.51
For each factor (SC, min LCC and CCC) the
0.5maximum, minimum,
0.52 mean
0.5and standard
deviation are shown.
CCC μ 0.68 0.64 0.62 0.61

σ 0.15 0.16 0.08 0.04

For each factor (SC, ECS/ECMS, LCC and CCC) the maximum, minimum, mean and standard
deviation are shown.
Sensors 2019, 19, 5214 12 of 16

4. Discussion and Conclusions

This article presented our study validating the low-cost sEMG system. The system was
developed using low-cost components and Matlab/Simulink software with the main idea of being
an economic alternative to commercial systems, keeping in mind the cutting-edge limitations of the
low-cost technology.
To reach that objective, an experiment to compare the output sEMG signal between a commercial
system and the proposed low-cost system was carried out. In that experiment each volunteer was
sensorized with both systems to register the sEMG signal of the rectus femoris (RF) while performing
some isometric and dynamic exercises. After that sEMG signal was used to carry out an analysis
following the explanation presented in Section 2.3 to validate the low-cost device. First, the signal
was synchronized, trimmed and denoised. Secondly, the MVC (maximal voluntary contraction) was
assessed to normalize the signal and finally some indicators to compare the similarity of both signals
were assessed.
Throughout the study some hypothesises were taken into account. The first one is related to the
relationship between the output sEMG signal and the exercise, making it possible to detect the exercise
with the output signal values. The second hypothesis states that the more difficulty to perform one
exercise, the higher output voltage is registered. Finally, the third hypothesis enclosed the possible
uses of the low-cost system for simple exercises, up to date.
The sEMG signal acquisition was done, as can be seen in Figure 2, by two different and isolated
computers. Because of that, both signals must be synchronized after the exercises. To that aim before
starting any exercise, each volunteer activated a trigger as matching point between both signals
(Figure 5). More information related to that can be found in the Section 2.3.1.
The exercises each volunteer had to perform were split into two groups, the isometric exercises and
dynamic exercises. The isometric exercises were used to obtain the MVC of each exercise to compare
the behaviour of the signal based on the muscle activation level. The dynamic exercises were used to
assess the four indicators to compare the sEMG signal from the low-cost and the commercial devices.
One of the first stages for the analysis was the signal denoising. It is important to highlight
that although the muscle is not stressed there could be still a low noise after being filtered. One of
the reasons why this may happen is due to the way the system is set up and the inherent noise of
the hardware employed [12]. Connections were done with braided cable, although they could be
improved in the future by means of welding terminals instead the pin system of the Arduino Mega
board. In addition, for a better reduction of the noise, it might be interesting to isolate the system,
although it is very difficult to remove the noise in an efficient way [13]. All those characteristics must
be studied and improved in future upgrades.
Results of the isometric exercises are plotted in Figure 6, on the right side the results from the
custom system and on the left side the results of the commercial system. As can be seen, in both
cases the tendency is the same, the output voltage of the signal increases according to the difficulty to
maintain the isometric position (second hypothesis). In addition, the output values for each exercise
define an average value, which makes it possible to identify the exercise, thereby fulfilling the first of
the hypotheses proposed. Aside from this, all the similarities suggest that the low-cost device works
in a satisfactory way, so it is possible to continue with the validation of the device (assessment of
the indicators). In addition, it is important to note that this research confirms the results related to
muscular behaviour in accordance with the subject studied [4].
The next stage of the study was the assessment of the indicators listed in the Section 2.3.3. They
were chosen to estimate the similarity between the two signals in a numerical way, so that the results
could validate the low-cost system. The Spearman’s coefficient calculated in all the exercises is above 0,
which means that the correlation between both signals is positive. In addition, results fluctuate between
a minimum of 0.51 and a maximum of 0.96, with an average above 0.60, which indicates similarity
between both signals. The next indicator was the energy ratio, the amount of energy along the exercise
of the low-cost device divided by the amount of energy along the exercise of the commercial system.
Sensors 2019, 19, 5214 13 of 16

This indicator varies from 0 to 1, where 0 is no similarities and 1 is 100% similarity between both signals.
Results from the experiment are from a minimum of 0.71 to a maximum of 0.98, with an average above
0.80. Those discrepancies are mainly due to the noise recorded by the custom equipment, since the
signal shape of the custom and commercial system is very similar, as can be seen in Figure 4. The third
factor is the linear correlation coefficient. It also varies between 0 and 1 (0, no correlation; and 1, perfect
correlation). In this case, the maximum, minimum and average values are very close to 1 (maximum 1,
minimum 0.98 and average 0.99), therefore linear correlation is almost 100%. The last indicator is the
cross-correlation coefficient (CCC). It moves between −1 to −1, where −1 means negative correlation
and 1 positive correlation. Results show a mean around 0.65, with a maximum of 0.87 that means a
good correlation between the signals.
All these findings indicate that there is a excellent agreement between the custom and the
commercial system. As a result, the third hypothesis is valid. The objective of this article has been
successfully achieved: Build a user-friendly system with low-cost parts.
In any case, although the achieved results were satisfactory, it is important to be aware of some
limitations that were found in the use of this low-cost system.
The first limitation, as mentioned above, is the problem with noise from both the assembly and
the equipment.
A second issue is related to the delay of the signal. The low-cost system has a slightly higher delay
than the commercial system, which could be a problem when tracking the signal in vivo. This delay is
lower than a second and must be studied in future analysis.
Also, it is important to have in mind that sEMG signal is very sensitive to the cleanliness of the
surface of positioning of the sensors.
One more factor is the sampling frequency, since the system is weighed down by the Arduino board.
Due to the fact that the system can be improved in future studies, more analysis and accurate
study of the limitations are going to be studied in the future.

Supplementary Materials: The following are available at http://www.mdpi.com/1424-8220/19/23/5214/s1.

Author Contributions: S.F.d.T., Y.W., L.R. and W.G. carried out the study conception and design. Funding’s
acquisitions were carried out thanks to V.D. and L.R. The acquisition of data was performed by S.F.d.T., Y.W.
S.F.d.T. worked with analysis and validation of data. S.F.d.T. wrote the manuscript. E.O., Y.W., L.R. and W.G.
provided critical feedback. Supervision of the tasks was developed by E.O. L.R. and W.G.
Funding: This research received no external funding.
Acknowledgments: Authors would like to thank the University of Manchester and to all the volunteers for
their contributions.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Hagberg, M. Work load and fatigue in repetitive arm elevations. Ergonomics 1981, 24, 543–555. [CrossRef]
[PubMed]
2. Petrofsky, J.S.; Glaser, R.M.; Phillips, C.A. Evaluation of the amplitude aud frequency components of the
surface emg as an index of muscle fatigue. Ergonomics 1982, 25, 213–223. [CrossRef] [PubMed]
3. Cifrek, M.; Medved, V.; Tonković, S.; Ostojić, S. Surface EMG based muscle fatigue evaluation in biomechanics.
Clin. Biomech. 2009, 24, 327–340. [CrossRef] [PubMed]
4. Fuentes del Toro, S.; Santos-Cuadros, S.; Olmeda, E.; Álvarez-Caldas, C.; Díaz, V.; San Román, J.L. Is the Use
of a Low-Cost sEMG Sensor Valid to Measure Muscle Fatigue? Sensors 2019, 19, 3204. [CrossRef] [PubMed]
5. Farina, D.; Pozzo, M.; Merlo, E.; Bottin, A.; Merletti, R. Assessment of Average Muscle Fiber Conduction
Velocity From Surface EMG Signals During Fatiguing Dynamic Contractions. IEEE Trans. Biomed. Eng.
2004, 51, 1383–1393. [CrossRef] [PubMed]
6. De Luca, C.J. Myoelectrical manifestations of localized muscular fatigue in humans. Crit. Rev. Biomed. Eng.
1984, 11, 251–279.
Sensors 2019, 19, 5214 14 of 16

7. Stolen, F.B.; De Luca, C.J.; De Luca, C.J. Frequency Parameters of the Myoelectric Signal as a Measure of
Muscle Conduction Velocity. IEEE Trans. Biomed. Eng. 1981, BME-28, 515–523. [CrossRef]
8. Bonato, P.; Roy, S.H.; Knaflitz, M.; de Luca, C.J. Time-frequency parameters of the surface myoelectric signal
for assessing muscle fatigue during cyclic dynamic contractions. IEEE Trans. Biomed. Eng. 2001, 48, 745–753.
[CrossRef]
9. Knaflitz, M.; Molinari, F. Assessment of muscle fatigue during biking. IEEE Trans. Neural Syst. Rehabil. Eng.
2003, 11, 17–23. [CrossRef]
10. Molinari, F.; Knaflitz, M.; Bonato, P.; Actis, M.V. Electrical Manifestations of Muscle Fatigue During Concentric
and Eccentric Isokinetic Knee Flexion-Extension Movements. IEEE Trans. Biomed. Eng. 2006, 53, 1309–1316.
[CrossRef]
11. Olmeda, E.; del Toro, S.F.; Garrosa, M.; Redondo, J.P.; Diaz, V. Clutch Pedal Sensorization and Evaluation of
the Main Parameters Related to Driver Posture. Sensors 2018, 18, 2797. [CrossRef] [PubMed]
12. Andrade, A.O.; Soares, A.B.; Nasuto, S.J.; Kyberd, P.J. EMG Decomposition and artefact removal.
In Computational Intelligence in Electromyography Analysis—A Perspective on Current Applications and Future
Challenges; InTech: London, UK, 2012; pp. 261–286.
13. Supuk, T.G.; Skelin, A.K.; Cic, M. Design, development and testing of a low-cost sEMG system and its use in
recording muscle activity in human gait. Sensors (Switz.) 2014, 14, 8235–8258. [CrossRef] [PubMed]
14. Heywood, S.; Pua, Y.H.; McClelland, J.; Geigle, P.; Rahmann, A.; Bower, K.; Clark, R. Low-cost
electromyography—Validation against a commercial system using both manual and automated activation
timing thresholds. J. Electromyogr. Kinesiol. 2018, 42, 74–80. [CrossRef] [PubMed]
15. Cheney, P.D.; Kenton, J.D.; Thompson, R.W.; McKiernan, B.J.; Lininger, R.E.; Trank, J.W. A low-cost,
multi-channel, EMG signal processing amplifier. J. Neurosci. Methods 1998, 79, 123–127. [CrossRef]
16. Tavakoli, M.; Benussi, C.; Lourenco, J.L. Single channel surface EMG control of advanced prosthetic hands:
A simple, low cost and efficient approach. Expert Syst. Appl. 2017, 79, 322–332. [CrossRef]
17. De Freitas, R.C.; Alves, R.; da Silva Filho, A.G.; de Souza, R.E.; Bezerra, B.L.D.; dos Santos, W.P.
Electromyography-controlled car: A proof of concept based on surface electromyography, Extreme Learning
Machines and low-cost open hardware. Comput. Electr. Eng. 2019, 73, 167–179. [CrossRef]
18. Castellini, C.; Gruppioni, E.; Davalli, A.; Sandini, G. Fine detection of grasp force and posture by amputees
via surface electromyography. J. Physiol. 2009, 103, 255–262. [CrossRef]
19. Nielsen, J.L.G.; Holmgaard, S.; Ning, J.; Englehart, K.B.; Farina, D.; Parker, P.A. Simultaneous and Proportional
Force Estimation for Multifunction Myoelectric Prostheses Using Mirrored Bilateral Training. IEEE Trans.
Biomed. Eng. 2011, 58, 681–688. [CrossRef]
20. Peerdeman, B.; Boere, D.; Witteveen, H.; in ’t Veld, R.H.; Hermens, H.; Stramigioli, S.; Rietman, H.; Veltink, P.;
Misra, S. Myoelectric forearm prostheses: State of the art from a user-centered perspective. J. Rehabil. Res. Dev.
2011, 48, 719–737. [CrossRef]
21. Schaeffer, E.L.; Liu, D.Y.; Guerin, J.; Ahn, M.; Lee, S.; Asaad, W.F. A low-cost solution for quantification of
movement during DBS surgery. J. Neurosci. Methods 2018, 303, 136–145. [CrossRef]
22. Angelova, S.; Ribagin, S.; Raikova, R.; Veneva, I. Power frequency spectrum analysis of surface EMG signals
of upper limb muscles during elbow flexion—A comparison between healthy subjects and stroke survivors.
J. Electromyogr. Kinesiol. 2018, 38, 7–16. [CrossRef] [PubMed]
23. Moritani, T.; Muro, M. Motor unit activity and surface electromyogram power spectrum during increasing
force of contraction. Eur. J. Appl. Physiol. Occup. Physiol. 1987, 56, 260–265. [CrossRef] [PubMed]
24. MathWorks Matlab 2017, R2017b; MathWorks: Natick, MA, USA, 2017.
25. Hips and Thighs. In Deep Tissue Massage Treat; Elsevier: Amsterdam, The Netherlands, 2013; pp. 134–157,
ISBN 9780323077590.
26. Merletti, R.; Roy, S. Myoelectric and Mechanical Manifestations of Muscle Fatigue in Voluntary Contractions.
J. Orthop. Sport. Phys. Ther. 2013, 24, 342–353. [CrossRef] [PubMed]
27. Garten, H.; Garten, H.M. quadriceps femoris. In Muscle Test Handbook; Elsevier: Amsterdam, The Netherlands,
2013; pp. 196–203.
28. Eckard, T.; Padua, D.; Mauntel, T.; Frank, B.; Pietrosimone, L.; Begalle, R.; Goto, S.; Clark, M.; Kucera, K.
Association between double-leg squat and single-leg squat performance and injury incidence among
incoming NCAA Division I athletes: A prospective cohort study. Phys. Ther. Sport 2018, 34, 192–200.
[CrossRef]
Sensors 2019, 19, 5214 15 of 16

29. Obokata, Y.; Nishi, K.; Kamatani, K. The dynamic characteristics of adolescent patients with spondylolysis
during forward lunge—the point aimed at trunk extension angle. Physiotherapy 2015, 101, 1115–1116.
[CrossRef]
30. Marchant, D.C.; Greig, M. Attentional focusing instructions influence quadriceps activity characteristics but
not force production during isokinetic knee extensions. Hum. Mov. Sci. 2017, 52, 67–73. [CrossRef]
31. Ancillao, A.; Galli, M.; Rigoldi, C.; Albertini, G. Linear correlation between fractal dimension of surface EMG
signal from Rectus Femoris and height of vertical jump. Chaos Solitons Fractals 2014, 66, 120–126. [CrossRef]
32. Miguel Gómez Palacios Exercise Drafts, 2019. Available online: https://bit.ly/2so5QmF (accessed on 28 November 2019).
33. Adam, A.; De Luca, C.J. Firing rates of motor units in human vastus lateralis muscle during fatiguing
isometric contractions. J. Appl. Physiol. 2005, 99, 268–280. [CrossRef]
34. Calder, K.M.; Stashuk, D.W.; McLean, L. Physiological characteristics of motor units in the brachioradialis
muscle across fatiguing low-level isometric contractions. J. Electromyogr. Kinesiol. 2008, 18, 2–15. [CrossRef]
35. Bickerstaffe, A.; van Dijk, J.P.; Beelen, A.; Zwarts, M.J.; Nollet, F. Loss of motor unit size and quadriceps
strength over 10 years in post-polio syndrome. Clin. Neurophysiol. 2014, 125, 1255–1260. [CrossRef]
36. Boonstra, T.W.; Daffertshofer, A.; van Ditshuizen, J.C.; van den Heuvel, M.R.C.; Hofman, C.;
Willigenburg, N.W.; Beek, P.J. Fatigue-related changes in motor-unit synchronization of quadriceps muscles
within and across legs. J. Electromyogr. Kinesiol. 2008, 18, 717–731. [CrossRef] [PubMed]
37. Advancer Technologies, LLC. Electromyography Sensor for Microcontroller Applications MyoWareTM Muscle
Sensor (AT-04-001) Datasheet; Advancer Technologies, LLC: Raleigh, CA, USA, 2015.
38. Hermie, J. Hermens Bart Freriks Roberto Merletti Dick Stegeman Joleen Blok Günter Rau Cathy
Disselhorst-Klug Göran. In European Recommendations for Surface Electromyography; SENIAM project,
Ed.; Roessingh Research and Development: Enschede, The Netherlands, 1999; ISBN 9075452-15-2.
39. Simsek, D. Different fatigue-resistant leg muscles and EMG response during whole-body vibration.
J. Electromyogr. Kinesiol. 2017, 37, 147–154. [CrossRef] [PubMed]
40. Carr, J.C.; Stock, M.S.; Hernandez, J.M.; Ortegon, J.R.; Mota, J.A. Additional insight into biarticular muscle
function: The influence of hip flexor fatigue on rectus femoris activity at the knee. J. Electromyogr. Kinesiol.
2018, 42, 36–43. [CrossRef] [PubMed]
41. Peña, G.G.; Consoni, L.J.; Wilian, M.; Siqueira, A.A.G. Feasibility of an optimal EMG-driven adaptive
impedance control applied to an active knee orthosis. Rob. Auton. Syst. 2019, 112, 98–108. [CrossRef]
42. Reeves, J.; Jones, R.; Liu, A.; Bent, L.; Nester, C. The between-day reliability of peroneus longus EMG during
walking. J. Biomech. 2019, 86, 243–246. [CrossRef]
43. Hermens, H.J.; Freriks, B.; Disselhosrt-Klug, C.; Rau, G. Development of recommendations for SEMG sensors
and sensor placement procedures. J. Electromyogr. Kinesiol. 2000, 10, 361–374. [CrossRef]
44. Limem, M.; Hamdi, M.A.; Maaref, M.A. Denoising uterine EMG signals using LMS and RLS adaptive
algorithms. In Proceedings of the 2016 2nd International Conference on Advanced Technologies for Signal
and Image Processing (ATSIP), Monastir, Tunisia, 21–23 March 2016; pp. 273–276.
45. Li, X.; Shin, H.; Zhou, P.; Niu, X.; Liu, J.; Rymer, W.Z. Power spectral analysis of surface electromyography
(EMG) at matched contraction levels of the first dorsal interosseous muscle in stroke survivors.
Clin. Neurophysiol. 2014, 125, 988–994. [CrossRef]
46. Cheng, K.; Yang, W. Using adaptive filter for extracting the surface diaphragmatic EMG signal. In Proceedings
of the 1992 14th Annual International Conference of the IEEE Engineering in Medicine and Biology Society,
Paris, France, 29 October–1 November 1992; Volume 6, pp. 2604–2605.
47. Salamea Palacios, C.R.; Luna Romero, S. Calibración Automática en Filtros Adaptativos para el Procesamiento
de Señales EMG. Rev. Iberoam. Automática e Informática Ind. 2019, 16, 232–237. [CrossRef]
48. Dufour, J.S.; Marras, W.S.; Knapik, G.G. An EMG-assisted model calibration technique that does not require
MVCs. J. Electromyogr. Kinesiol. 2013, 23, 608–613. [CrossRef]
49. Delsys Inc. EMGworks Analysis; Delsys Inc.: Natick, MA, USA, 2019.
50. Microsoft Microsoft Office 2019; Microsoft Press: Washington, DC, USA, 2018.
51. Wei, Y.; Wei, G.; Ren, L. Predict Afferent Tactile Neural Signal for Artificial Nerve Based on Finite Element
Human Hand Model. In Intelligent Robotics and Applications; Yu, H., Liu, J., Liu, L., Ju, Z., Liu, Y., Zhou, D., Eds.;
Springer International Publishing: Cham, Switzerland, 2019; pp. 129–140.
Sensors 2019, 19, 5214 16 of 16

52. Kwon, D.-Y.; Park, B.K.; Kim, J.W.; Eom, G.-M.; Hong, J.; Koh, S.-B.; Park, K.-W. Quantitative
Electromyographic Analysis of Reaction Time to External Auditory Stimuli in Drug-Naïve Parkinson’s
Disease. Parkinsons Dis. 2014, 2014, 848035.
53. Ateş, F.; Davies, B.L.; Chopra, S.; Coleman-Wood, K.; Litchy, W.J.; Kaufman, K.R. Intramuscular Pressure of
Tibialis Anterior Reflects Ankle Torque but Does Not Follow Joint Angle-Torque Relationship. Front. Physiol.
2018, 9, 22. [CrossRef] [PubMed]
54. Power, S. Signal Energy and Correlation. In Continuous-Time Signals; Springer: Amsterdam, The Netherlands,
1989; Volume 2, pp. 202–254.
55. Harrison, L.M.; Ironton, R.; Stephens, J.A. Cross-correlation analysis of multi-unit EMG recordings in man.
J. Neurosci. Methods 1991, 40, 171–179. [CrossRef]
56. Winter, D.A.; Fuglevand, A.J.; Archer, S.E. Crosstalk in surface electromyography: Theoretical and practical
estimates. J. Electromyogr. Kinesiol. 1994, 4, 15–26. [CrossRef]
57. Luca, C.J. De The Use of Surface Electromyography. Appl. Biomech. 1997, 13, 135–163. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).