See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331451812

Validity and reliability of the WIMU ® system to measure barbell velocity

during the half-squat exercise

Article  in  Proceedings of the Institution of Mechanical Engineers Part P Journal of Sports Engineering and Technology · March 2019
DOI: 10.1177/1754337119831890

CITATIONS READS

4 582

5 authors, including:

Felipe García Pinillos Pedro Ángel Latorre Román

University of Granada Universidad de Jaén
161 PUBLICATIONS   1,274 CITATIONS    243 PUBLICATIONS   1,681 CITATIONS   

SEE PROFILE SEE PROFILE

Carlos Balsalobre-Fernández J.A. Párraga Montilla

Universidad Autónoma de Madrid Universidad de Jaén
70 PUBLICATIONS   1,344 CITATIONS    74 PUBLICATIONS   330 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Velocity based training View project

Small-sided games in football players View project

All content following this page was uploaded by Felipe García Pinillos on 05 March 2019.

The user has requested enhancement of the downloaded file.

Original Article

Proc IMechE Part P:

J Sports Engineering and Technology

Validity and reliability of the WIMUÒ 1–8

Ó IMechE 2019
Article reuse guidelines:
system to measure barbell velocity sagepub.com/journals-permissions
DOI: 10.1177/1754337119831890

during the half-squat exercise journals.sagepub.com/home/pip

Felipe Garcı́a-Pinillos1 , Pedro A Latorre-Román2,

Fernando Valdivieso-Ruano2, Carlos Balsalobre-Fernández3,4 and
Juan A Párraga-Montilla2

Abstract
This study aimed at determining the reliability and concurrent validity of the WIMUÒ system when measuring barbell
velocity during the half-squat exercise by comparing data with the gold standard. A total of 19 male competitive powerlif-
ters performed an incremental loading test using the half-squat exercise. The mean velocity, mean propulsive velocity
and maximum velocity of all repetitions were recorded through both WIMU and T-Force systems. As a measure of relia-
bility, coefficient of variations ranged from 6%–17% and standard error of means ranged from 0.02–0.11 m/s, showing
very close reliability of data from both devices. Validity, in terms of coefficient of correlations and pairwise comparisons,
was also tested. Except for some relative loads, the Pearson correlation analysis revealed significant correlations
between both devices for mean velocity, mean propulsive velocity and maximum velocity (r . 0.6, p \ 0.05). The mean
velocity, mean propulsive velocity and maximum velocity were underestimated for the WIMU system compared to T-
Force data at some points of the load–velocity relationship. The linear regression models performed revealed a strong
load–velocity relationship in the half-squat exercise for each individual using mean velocity, mean propulsive velocity and
maximum velocity, regardless of the instrument used (R2 . 0.77 in all cases). Bland–Altman plots revealed low systematic
bias (40.06 m s21) and random error (40.07 m s21) for the mean velocity and mean propulsive velocity obtained from the
WIMU system as compared to the T-Force, while the maximum velocity resulted in an underestimation by the WIMU sys-
tem (–0.16 m s21) as compared to the linear position transducer system. The results indicate that the WIMU system is a
reliable tool for tracking barbell velocity in the half squat, but these data also reveal some limitations regarding its concurrent
validity as compared to the gold standard, with velocity measures slightly underestimated in the tested conditions.

Keywords
velocity-based training, accelerometer, inertial sensor, resistance training

Date received: 22 August 2018; accepted: 26 January 2019

Introduction factors affect the accuracy of the prediction. Likewise,

the type of data analysis (generalized vs individualized
Assessments of velocity, force and power are often
employed to monitor training induced adaptations.1,2
Velocity measurement has been demonstrated to be a 1
Department of Physical Education, Sports and Recreation, Universidad
non-invasive methodology to estimate the one- de La Frontera, Temuco, Chile
2
repetition maximum (1RM, i.e. the maximum load that Department of Corporal Expression, University of Jaen, Jaén, Spain
3
can be lifted only one time) of the participants based Department of Physical Education, Sport and Human Movement,
Universidad Autónoma de Madrid, Madrid, Spain
on the well-known load–velocity relationship.1,3–5 As 4
LFE Research Group, Department of Health and Human Performance,
indicated by few previous studies,1,3,4 lifting velocity Faculty of Physical Activity and Sport Science-INEF, Universidad
and load (i.e. load–velocity relationship) are significantly Politécnica de Madrid, Madrid, Spain
correlated. Therefore, the 1RM can be estimated based on
velocity data without performing an actual 1RM test.1,3–5 Corresponding author:
Felipe Garcı́a-Pinillos, Department of Physical Education, Sports and
Nevertheless, the load–velocity relationship seems to be Recreation, Universidad de La Frontera, Calle Uruguay, 1980, Temuco,
exercise-dependent and specific to the execution mode 4811230, Chile.
(e.g. eccentric-concentric vs concentric-only),6,7 so those Email: [email protected]
2 Proc IMechE Part P: J Sports Engineering and Technology 00(0)

load–velocity relationship) seems to be an influencing fac- compared to the T-force linear velocity transducer for
tor on the accuracy of the 1RM prediction. For example, measuring barbell velocity on the half-squat exercise
Pérez-Castilla et al.6 reported different R2 values for gen- during an incremental protocol.
eralized and individualized load–velocity relationship dur-
ing the concentric-only half-squat exercise (0.920 vs 0.976,
respectively).
Participants
For elite athletes, changes in these measures can be A total of 19 male, national competitive powerlifters
minor, yet significant. Small differences in velocity (age range: 19–30 years; mean: 23.7 6 2.8 years old;
measures during a lift (i.e. \ 0.1 m/s) could represent body height: 1.77 6 0.06 m; body mass: 79.9 6 9.5 kg;
important variations in training intensity.1,3 For exam- half-squat 1RM: 147.5 6 27.1 kg; half-squat 1RM rela-
ple, Conceicxão et al.3 reported changes in barbell velo- tive to body mass: 1.85 6 0.25 kg kg21) participated in
city during a half squat (i.e. ;0.06 m/s) with 5% this study. Selection criteria included having at least
increases in the load (in terms of %1RM). As a conse- 3 years of experience in powerlifting training, able to lift
quence, equipment used to monitor changes in velocity a load greater than their body mass in the half-squat
should be precise. In a laboratory-based environment, exercise. Participants reported no chronic diseases or
linear position transducers (LPTs) are often used to recent injuries that could compromise test performance.
accurately measure velocity. Technologies such as They were instructed to avoid any strenuous exercise
accelerometers or photogrammetry have been also used two days prior to each testing session. Participants were
to track barbell velocity,8–14 but LPT systems are con- informed of the study procedures and signed a written
sidered the gold standard by many researchers. informed consent form prior to participating in the
Importantly, laboratory-based equipment (i.e. LPT sys- study. The study protocol adhered to the tenets of the
tems) is limited due to the large expense, which, in most Declaration of Helsinki and was approved by the
cases, is more expensive than accelerometers and iner- Institutional Review Board.
tial sensors. The presence of multiple cables in LPT sys-
tems hinders its use during measurements (i.e. wireless Procedures
measurement units seem to be easier to handle and suit-
able for portable applications on the field). Subjects arrived at the laboratory in a well-rested con-
Consequently, several field-based devices, including dition at the start of each testing session. The warm-up
accelerometers and inertial sensors (accelerometer plus routine was created based on the preferences of the par-
gyroscope) have been invented to overcome these lim- ticipants and consisted of jogging (5 min at a comforta-
itations.8–14 In addition, an increased interest in ble speed), foam rolling (3–4 min focused on lower back
cheaper, wireless measurement tools to measure velo- and hip extensors and flexors), dynamic stretching and
city without direct attachments, which makes the mea- lower-body joint mobilization exercises (thoracic, hip
surement process more affordable and easier. and ankle mobility for 3–4 min), followed by one set of
The reliability and validity of a barbell velocity analy- five repetitions with an absolute load of 18 kg in the
sis system are essential to determine whether results are half-squat exercise. Thereafter, subjects performed an
due to changes in mechanical properties or are simply incremental loading test using the half-squat exercise.
systematic measurement errors. In that context, a previ- The concentric-only half-squat test was performed.
ous study15 focused on determining the reliability and Subjects first flexed their knees to 90° in a continuous
validity of the WIMUÒ system for measuring barbell and controlled manner. They maintained this position
for 1.5 s, and immediately afterwards, they performed a
velocity during resistance exercises (i.e. back squat using
purely concentric action at maximal intended velocity
Smith machine). Specifically, that work determined the
without lifting their toes off the ground. The duration
reliability for mean velocity (MV) measurements at 40%
of the eccentric and isometric phases was paced by
and 80% 1RM, as well as to compare the results from a
auditory signals. Subjects initiated the movement in a
linear encoder in terms of concurrent validity. Therefore,
fully extended position, feet approximately shoulder-
reliability and validity data for other mechanical vari-
width apart and the barbell held across the top of the
ables provided by the WIMU system (i.e. mean propul-
shoulders and upper back. The barbell was required to
sive velocity (MPV) and maximum velocity (Vmax) are
be in constant contact with the subjects’ shoulders and
still unknown. The aim of this study is to determine the
upper-back through the whole execution. The 90° knee
reliability and concurrent validity of the WIMU system
angle was individually measured with a manual goni-
for measuring barbell velocity on the half-squat exercise,
ometer. To ensure the reproducibility of the 90° knee
by comparing data with the gold standard LPT system.
angle, a tripod adjustable with a telemetric photocell
(Microgate, Bolzano, Italy) was placed on one side of
the bar.6 The telemetric photocell emitted an acoustic
Methods
signal when the bar crossed the depth linked to the 90°
With the introduction of new wireless devices, establish- knee angle for each subject. The height of the tripod
ment of their reliability and validity are essential before was the same over the entire testing session. For each
practical use. In this study, the WIMU system was repetition, subjects received real-time velocity
Garcı́a-Pinillos et al. 3

performance feedback to encourage them to give maxi- Statistical analysis

mal effort. Trained spotters were present, and lifting Descriptive statistics are represented as mean standard
belts were used to ensure safety. deviation (SD). Tests of normal distribution and homo-
During the testing session, the initial load was set at geneity, determined by the Shapiro–Wilk and Levene’s
18 kg for all subjects and progressively increased in 20 kg test, respectively, were conducted on all data before
increments until the MPV attained was \ 0.6 m/s. analysis. In order to make comparisons easier, absolute
Then, load increments were individualized based on loads (i.e. kg) were converted into relative loads (i.e.
feedback from the participants and the protocol was %1RM). For that purpose, the best performance dur-
continued until no further weight could be lifted. Two ing the protocol, in terms of higher load lifted, was con-
attempts per load were executed with a recovery period sidered as 1RM (100%), and the rest of loads were
of 1 min in-between. Rests between sets were set to 3 min calculated as percentages of 1RM. Coefficient of varia-
for the lighter loads (MPV ø 1.00 m/s) and 5 min for tion (CV, %) and standard error of the mean (SEM)
the medium loads (MPV \ 1.00 m/s). Only the repeti- were calculated as a measure of absolute reliability
tion with the highest MPV at each load (measured (within-subject variation (CV = SD*mean*100) and
through the T-Force system) was selected for further SD of a sampling distribution (SEM = SD*O(1-relia-
analysis,1,6,16 apart from reliability measures which bility)), respectively).18,19 For that purpose, the two
required two attempts per load. The average MPV for attempts per load were used, and CV and SEM were
the heaviest load used in the incremental loading test calculated on an individual basis. To determine concur-
was 0.43 6 0.06 m/s. The average number of incremental rent validity, a Pearson correlation analysis was also
loads tested was 8.6 6 1.6. performed between T-Force and WIMU data. The fol-
lowing criteria were adopted to interpret the magnitude
of correlations between measurement variables: \ 0.1
Materials and testing (trivial), 0.1–0.3 (small), 0.3–0.5 (moderate), 0.5–0.7
Height (Seca 202, Seca Ltd., Hamburg, Germany) and (large), 0.7–0.9 (very large) and 0.9–1.0 (almost per-
body mass (Tanita BC-418 segmental, Tokyo, Japan) fect).20 Pairwise comparisons of means (t-test) were
were assessed in the first testing session. A Smith also conducted between data (MV, MPV and Vmax)
machine (Technogym, Barcelona, Spain) coupled with from the two devices at different loads (in terms of
an LPT system (T-Force Dynamic Measurement %1RM), and linear regressions were used to analyze
System; Ergotech, Murcia, Spain), which sampled the the load–velocity relationship for each subject. Finally,
velocity of the barbell at a frequency of 1,000 Hz, was Bland–Altman plots (i.e. limits of agreement method,
used in the testing session.17 In addition, an inertial mean difference 6 1.96 SD)21 were constructed to
device (WIMU system, RealTrack systems, Almerı́a, examine the presence of systematic and proportional
Spain) was attached to the bar. This device contained bias between the measured (i.e. T-Force system) and
four three-dimensional (3D) accelerometers, among estimated values of mechanical variables (i.e. WIMU
other sensors (three 3D gyroscopes and magnet- system). The level of significance used was p \ 0.05.
ometers), that detect and measure movement using a Data analysis was performed using SPSS (version 23,
micro-electromechanical system with an adjustable SPSS Inc, Chicago, IL).
sample frequency of 10–1000 Hz. The sample frequency
used in this study was 1000 Hz. To measure velocity, Results
the acceleration measured by triaxial accelerometers
was used. From the acceleration, the device estimates Reliability
velocity of each repetition in both the concentric and Table 1 shows the %CV (as a measure of absolute
eccentric phases (only concentric data were used for the reliability) of mechanical parameters at different loads
subsequent analysis). Each trial was downloaded using from both systems. For the LPT system, %CV was
the manufacturer’s software S PROä (RealTrack sys- 6.8–16.8% (MV), 8.3–17.3% (MPV) and 5.7–9.9%
tems, Almerı́a, Spain). (Vmax), whereas for the WIMU system, %CV ranged
The MV (i.e. the average velocity from the start of between 6.2–13.8% (MV), 8.2–16.8% (MPV) and 7.2–
the concentric phase until the bar reaches the maximum 16.8% (Vmax). In addition, the SEM is provided in
height), MPV (i.e. the average velocity from the start of Table 2.
the concentric phase until the acceleration of the bar is
less than gravity (–9.81 m/s22)) and Vmax (i.e. the maxi-
mum instantaneous velocity value reached during the Validity
concentric phase) of all repetitions were recorded.16 The Pearson correlation analysis is shown in Table 3.
Only the repetition with the highest MPV of each load- The MV values from both devices showed significant
ing condition (measured through the T-Force system) correlations over the protocol (r ø 0.646, p \ 0.05),
was used for statistical analyses. Before all attempts, except for 20% and 70% 1RM (r 4 0.520, p ø 0.05).
both devices (T-Force and WIMU systems) were cali- The MPV values showed significant large and very
brated according to the manufacturer’s instructions. larges correlations (r ø 0.628, p \ 0.05) at every
4 Proc IMechE Part P: J Sports Engineering and Technology 00(0)

Table 1. Coefficient of variation (%) of variables related to the Table 2. Standard error of mean (SEM) of variables related to
measurement of barbell velocity (MV, MPV and Vmax) in both the the measurement of barbell velocity (MV, MPV and Vmax) in both
linear position transducer (LPT) and the WIMU system. the linear position transducer (LPT) and the WIMU system.

%1RM MV (m/s) MPV (m/s) Vmax (m/s) %1RM MV (m/s) MPV (m/s) Vmax (m/s)
LPT WIMU LPT WIMU LPT WIMU LPT WIMU LPT WIMU LPT WIMU

10 8.60 9.02 11.28 11.69 9.18 11.76 10 0.03 0.03 0.04 0.04 0.06 0.07
20 6.76 6.19 8.29 8.14 8.17 8.45 20 0.02 0.02 0.03 0.03 0.05 0.05
30 11.86 11.77 14.39 12.44 7.42 14.44 30 0.11 0.11 0.04 0.03 0.04 0.06
40 9.95 7.90 10.87 8.32 6.64 7.48 40 0.09 0.06 0.03 0.02 0.03 0.03
50 9.06 7.86 11.21 8.99 6.49 7.52 50 0.07 0.06 0.02 0.02 0.03 0.03
60 9.27 10.41 13.81 11.23 9.91 13.62 60 0.02 0.02 0.03 0.02 0.04 0.05
70 9.50 13.82 12.27 16.75 8.68 16.80 70 0.02 0.02 0.02 0.03 0.03 0.05
80 9.09 12.04 9.07 13.23 5.67 7.18 80 0.02 0.02 0.02 0.02 0.02 0.03
90 11.02 12.62 11.00 16.47 7.73 11.77 90 0.05 0.06 0.02 0.02 0.02 0.03
100 16.77 13.27 17.26 14.21 9.79 15.66 100 0.02 0.01 0.02 0.01 0.02 0.03

MV: the average velocity from the start of the concentric phase until the MV: the average velocity from the start of the concentric phase until the
bar reaches the maximum height; MPV: the average velocity from the bar reaches the maximum height; MPV: the average velocity from the
start of the concentric phase until the acceleration of the bar is lower start of the concentric phase until the acceleration of the bar is lower
than gravity (–9.81 m s22)); Vmax: the maximum instantaneous velocity than gravity (–9.81 m s22)); Vmax: the maximum instantaneous velocity
value reached during the concentric phase; 1RM: the maximum load that value reached during the concentric phase; 1RM: the maximum load that
can be lifted only one time; LPT: linear position transducers. can be lifted only one time; LPT: linear position transducer.

Table 3. Pearson correlation between variables related to the measurement of barbell velocity (MV, MPV and Vmax) from the linear
position transducer (LPT) and the WIMU system over an incremental test.

%1RM 10 20 30 40 50 60 70 80 90 100

MV (m/s) 0.865** 0.520 0.696** 0.877** 0.760** 0.646* 0.419 0.662* 0.739** 0.687**
MPV (m/s) 0.898** 0.398 0.813** 0.882** 0.823** 0.645* 0.628* 0.632* 0.717** 0.685**
Vmax (m/s) 0.971** 0.773** 0.196 0.842** 0.908** 0.729* 0.819** 0.498 0.742** 0.861**

MV: the average velocity from the start of the concentric phase until the bar reaches the maximum height; MPV: the average velocity from the start
of the concentric phase until the acceleration of the bar is lower than gravity (–9.81 m s22)); Vmax: the maximum instantaneous velocity value reached
during the concentric phase; 1RM: the maximum load that can be lifted only one time; LPT: linear position transducer.
*p \ 0.05; **p \ 0.01; ***p \ 0.001.

%1RM except for 20% (r = 0.398, p ø 0.05). The velocity relationship (R2 . 0.75 in all cases) exists in
Vmax values from both devices showed significant cor- the half-squat exercise for each individual using MV,
relations over the protocol (r ø 0.729, p \ 0.05), MPV and Vmax, no matter which instrument was used.
except for 30% and 80% 1RM (r 4 0.498, p ø 0.05). Specifically, similar R2 values, although always lower
A paired t-test (Figure 1) demonstrated significant for the WIMU system, were obtained with the T-Force
differences (p \ 0.05) between data from both devices and the WIMU for load-MV (LPT: R2 = 0.89;
at similar %1RM. The MV was underestimated for WIMU: R2 = 0.86) and load-MPV relationship (LPT:
WIMU system compared to T-Force data (p \ 0.05) R2 = 0.86; WIMU: R2 = 0.85), with some differences
with no differences at 10% and 60–80%1RM. The in the load–Vmax relationship (LPT: R2 = 0.87;
MPV was also underestimated for the WIMU system WIMU: R2 = 0.77).
compared to T-Force data (p \ 0.05) with no differ- Through Bland–Altman plots, Figure 3 shows the
ences at 30% and 100%1RM. Likewise, Vmax was differences between the two devices (systematic bias
underestimated for the WIMU system over the entire and random error) and the degree of agreement
protocol (p \ 0.01). between the two systems (95% limits of agreement).
Figure 2 shows the load–velocity relationship mea- These plots revealed low systematic bias (40.06 m s21)
sured with both the LPT and the WIMU system over and random error (40.07 m s21) for the MV and MPV
the incremental test (in terms of %1RM). Velocity obtained from the WIMU system as compared to the
measures (MV, MPV and Vmax) were plotted against T-Force, while the Vmax was underestimated by the
the load lifted. The results showed that strong load– WIMU system (–0.16 m s21) as compared to the LPT
Garcı́a-Pinillos et al. 5

Figure 1. Paired t-test between velocity data (MV, MPV and Vmax) from both devices (LPT vs WIMU systems) at similar loads
(%1RM).
MV: mean velocity; MPV: mean propulsive velocity; Vmax: maximum velocity; LPT: linear position transducer; %1RM: percentage of one-repetition
maximum.
*p \ 0.05, **p \ 0.01, ***p \ 0.01.

Figure 2. Load–velocity relationship measured with both the LPT and the WIMU systems over the incremental test.
LPT: linear position transducer.
6 Proc IMechE Part P: J Sports Engineering and Technology 00(0)

Figure 3. Bland–Altman plots for the measurement of mechanical variables for both systems (T-Force vs WIMU): (a) mean velocity,
(b) mean propulsive velocity, and (c) maximum velocity.
The plot includes the mean difference (dotted line) and 95% limits of agreement (dashed lined), along with the regression line (solid line).

system. Limits of agreement showed a similar width, variables from both devices were large and very large20
though slightly wider for the Vmax as compared to the for MV, MPV and Vmax over the entire protocol,
MV and MPV. except MPV and Vmax values at light loads (i.e. ;20%–
30% 1RM). The linear regression model and Bland–
Altman plots support the validity analysis. The two
Discussion regression lines virtually overlap when plotted on the
This study aimed to determine the reliability and con- same graph, showing a solid load–velocity relationship
current validity of the WIMU system compared to the (0–80 \ R2 . 0.90), with lower R2 values for the
T-Force system for measuring barbell velocity on the WIMU system, whereas the Bland–Altman plots report
half-squat exercise. As a measure of within-subject var- an assumable systematic bias for MV and MPV, but
iation and SD of a sampling distribution, the CV ran- too high for Vmax.
ged from ;6–17% and the SEM ranged from ;0.02– Since LPT systems are considered the gold standard
0.11 m/s, showing very close reliability data from both for measuring barbell velocity,1,4,8–10,22 the results
devices. Velocity measures, in terms of MV, MPV and obtained point to the WIMU system as a reliable tool
Vmax, were very similar from both devices. However, for tracking barbell velocity in the half squat, with simi-
there were some differences between the two devices as lar CV and SEM to those reported by the T-Force sys-
revealed by the paired samples t-test, which showed tem, but these data also suggest some limitations
that the velocities measured with the WIMU tended to regarding its concurrent validity, as compared to the
be slightly lower than those measured with the LPT. gold standard, with velocity measures slightly underes-
To reinforce the concurrent validity analysis, a Pearson timated in the tested conditions. The LPT system used
correlation analysis was conducted for each variable. in this study directly measured the vertical displacement
The magnitude of correlations between measurement velocity of its cable through a sensor attached to the
Garcı́a-Pinillos et al. 7

barbell. Compared to other LPTs, this system trans- Previous studies have assessed the reliability and
duces electrical signals, rather than differentiates cable validity of different low-cost and wireless devices, as
position over time,1,22 showing a very high accuracy.17 compared to the T-Force system, for tracking barbell
As mentioned in the introduction, a previous study15 velocity.8–14,24 Specifically, the PUSH wearable device
analyzed the reliability and validity of the WIMU sys- (https://www.trainwithpush.com) showed high levels of
tem for measuring barbell velocity during the back validity and reliability in comparison with a LPT
squat, but in that case, only the MV measurements at (r . 0.9) for the measurement of movement velocity
loads equivalent to 40% and 80% were taken into con- during back squats.9 That data slightly decreased in a
sideration. The results from that study are similar to more recent work during the same exercise (i.e. back
the results from this study, indicating that the WIMU squat, r = 0.86).10 Balsalobre-Fernández et al.10 also
system is a reliable tool, even though reliability was provide data about other wearable sensors (BEAST
examined through test–retest analysis, whereas this sensor, https://www.thisisbeast.com/en), which showed
study determined absolute reliability. Regarding the superior values of correlation (r . 0.96) than the LPT
validity analysis, the results obtained in this study are system. These differences could be due to technical
similar to those reported by Muyor et al.,15 with barbell issues (accuracy differences), but some points need to
velocity slightly underestimated by the WIMU system be addressed. First, the sample frequency differed.
as compared to the T-Force system. In addition, this Whereas, the WIMU system was tested at 1000 Hz, the
study determined the dynamic of velocity-based vari- PUSH and BEAST sensors were tested at 200 and
ables (i.e. MV, MPV and Vmax) over an incremental 50 Hz, respectively. Second, each wearable sensor is
protocol and analyzed the load–velocity relationship designed to be placed on different body parts. The
and the accuracy of 1RM estimations based on data BEAST sensor is designed to be placed in a wristband
reported from these two devices (WIMU vs T-Force or directly attached to the barbell. The PUSH band is
systems). meant to be placed just below the elbow of the subject.
Based on the evidence supporting the importance of The WIMU system was attached to the barbell. Finally,
measuring movement velocity to monitor training the type of resistance exercise tested varied (in this case,
adaptations1,2 and the accuracy of velocity measure- full squat vs half squat). The lack of methodological
ments to estimate the 1RM (taking into account the consistency makes comparisons difficult.
exercise, execution mode and data analysis),1,3–7 the
measurement of the load–velocity relationship (based Conclusion
on a linear regression model) seems to be well justified.
The load–velocity profile describes the ability of the The results from this study demonstrate an acceptable
subject to produce velocity at increasing intensities reliability of the WIMU system compared with an LPT
(normally, during an incremental test). The load– system, but these data also suggest some limitations
velocity profile has been shown to be an accurate regarding its concurrent validity as compared to the
method to estimate 1RM,4,23 even though it is exercise- gold standard, with velocity measures slightly underes-
dependent and specific to the execution mode.6,7 timated in the test conditions.
Previous studies have analyzed the load–velocity pro- From a practical point of view, scientists and coa-
files of different exercises such as the full squat,3 half ches should know that both devices showed an ade-
squat5,6 and bench press1,7 with high coefficients of quate reliability for barbell velocity assessment.
determination reported when data analysis was Therefore, velocity-based variables reported from these
individualized. devices can be compared over time if using the same
This study shows that load–velocity profiles derived device. However, the clients also should be aware about
from the velocities measured with LPT versus WIMU the limitations of comparing data from these two
system reported similar results for MV and MPV devices with the WIMU system useful on an individual
(almost identical R2 values), with varying results for basis.
Vmax (R2 = 0.87 and 0.77 for LPT and WIMU, respec-
tively). Despite the WIMU system showing homoge- Declaration of conflicting interests
neous differences across the entire protocol (increasing
The author(s) declared no potential conflicts of interest
loads in terms of %1RM) compared to the LPT, the with respect to the research, authorship, and/or publi-
authors strongly recommend not to use LPT and cation of this article.
WIMU systems interchangeably. As reported by previ-
ous studies,1,3 small differences in velocity measures
during a lift (i.e. \ 0.1 m/s) could represent important Funding
variations in training intensity, causing training load The author(s) received no financial support for the
and targets to be overestimated or underestimated. research, authorship, and/or publication of this article.
8 Proc IMechE Part P: J Sports Engineering and Technology 00(0)

ORCID iD Res 2011; 25(8): 2293–2297, http://www.ncbi.nlm.nih.

gov/pubmed/21747293
Felipe Garcı́a-Pinillos https://orcid.org/0000-0002-
13. Sañudo B, Rueda D, Del Pozo-Cruz B, et al. Validation
7518-8234
of a video analysis software package for quantifying
movement velocity in resistance exercises. J Strength
References Cond Res 2016; 30(10): 2934–2941.
14. Banyard HG, Nosaka K, Sato K, et al. Validity of vari-
1. González-Badillo JJ and Sánchez-Medina L. Movement
ous methods for determining velocity, force, and power in
velocity as a measure of loading intensity in resistance
the back squat. Int J Sports Physiol Perform 2017; 12(9):
training. Int J Sports Med 2010; 31(5): 347–352.
1170–1176.
2. Suchomel TJ, Nimphius S and Stone MH. The impor-
15. Muyor JM, Granero-Gil P and Pino-Ortega J. Reliability
tance of muscular strength in athletic performance. Sport
and validity of a new accelerometer (WimuÒ) system for
Med 2016; 46: 1419–1449.
3. Conceicxão F, Fernandes J, Lewis M, et al. Movement measuring velocity during resistance exercises. Proc
velocity as a measure of exercise intensity in three lower IMechE, Part P: J Sports Engineering and Technology
limb exercises. J Sports Sci 2016; 34(12): 1099–1106. 2018; 232(3): 218–224.
4. Jidovtseff B, Harris NK, Crielaard JM, et al. Using the 16. Garcı́a-Ramos A, Pestaña-Melero F, Pérez-Castilla A, et
load-velocity relationship for 1RM prediction. J Strength al. Mean velocity vs. mean propulsive velocity vs. peak
Cond Res 2011; 25(1): 267–270. velocity: which variable determines bench press relative
5. Loturco I, Pereira LA, Cal Abad CC, et al. Using bar load with higher reliability? J Strength Cond Res 2018; 32:
velocity to predict maximum dynamic strength in the 1273–1279.
half-squat exercise. Int J Sports Physiol Perform 2016; 17. Garnacho-Castaño MV, López-Lastra S and Maté-
11(5): 697–700. Muñoz JL. Reliability and validity assessment of a linear
6. Pérez-Castilla A, Garcı́a-Ramos A, Padial P, et al. Load- position transducer. J Sport Sci Med 2014; 14(1): 128–
velocity relationship in variations of the half-squat exer- 136.
cise: influence of execution technique. J Strength Cond 18. Atkinson G and Nevill AM. Statistical methods for asses-
Res. Epub ahead of print 4 April 2018. DOI: 10.1519/ sing measurement error (reliability) in variables relevant
JSC.0000000000002072. to sports medicine. Sport Med 1998; 26(4): 217–238.
7. Sánchez-Medina L, González-Badillo JJ, Pérez CE, et al. 19. Hopkins WG. Measures of reliability in sports medicine
Velocity- and power-load relationships of the bench pull and science. Sport Med 2000; 30(1): 1–15.
vs bench press exercises. Int J Sports Med 2014; 35(3): 20. Hopkins WG, Marshall SW, Batterham AM, et al. Pro-
209–216. gressive statistics for studies in sports medicine and exer-
8. Balsalobre-Fernández C, Marchante D, Muñoz-López cise science. Med Sci Sports Exerc 2009; 41(1): 3–13.
M, et al. Validity and reliability of a novel iPhone app 21. Bland JM and Altman DG. Comparing methods of mea-
for the measurement of barbell velocity and 1RM on the surement: why plotting difference against standard
bench-press exercise. J Sports Sci 2018; 36: 64–70. method is misleading. Lancet 1995; 346(8982): 1085–
9. Balsalobre-Fernández C, Kuzdub M, Poveda-Ortiz P, et 1087.
al. Validity and reliability of the PUSH wearable device 22. Cormie P, Deane R and McBride JM. Methodological
to measure movement velocity during the back squat concerns for determining power output in the jump squat.
exercise. J Strength Cond Res 2016; 30(7): 1968–1974. J Strength Cond Res 2007; 21(2): 424–430.
10. Balsalobre-Fernández C, Marchante D, Baz-Valle E, et 23. Picerno P, Iannetta D, Comotto S, et al. 1RM prediction:
al. Analysis of wearable and smartphone-based technolo- a novel methodology based on the force–velocity and
gies for the measurement of barbell velocity in different load–velocity relationships. Eur J Appl Physiol 2016;
resistance training exercises. Front Physiol 2017; 8: 649. 116(10): 2035–2043.
11. Crewther BT, Kilduff LP, Cunningham DJ, et al. Vali- 24. Garcı́a-Ramos A, Pérez-Castilla A and Martı́n F. Relia-
dating two systems for estimating force and power. Int J bility and concurrent validity of the Velowin optoelectro-
Sports Med 2011; 32(4): 254–258. nic system to measure movement velocity during the free-
12. Comstock BA, Solomon-Hill G, Flanagan SD, et al. weight back squat. Int J Sports Sci Coa 2018; 13: 737–
Validity of the MyotestÒ in measuring force and power 742.
production in the squat and bench press. J Strength Cond

View publication stats