Gait & Posture 58 (2017) 154158

Contents lists available at ScienceDirect

Gait & Posture

journal homepage: www.elsevier.com/locate/gaitpost

Full length article

Hamstring and calf muscle activation as a function of bodyweight support MARK

during treadmill running in ACL reconstructed athletes

Clint Hansena, , Einar Einarsonb, Athol Thomsonb, Rodney Whiteleyb, Erik Witvrouwc
a
Department of Neurology, UKSH, Kiel University Christian-Albrechts-Universitt zu Kiel, Kiel, Germany
b
Aspetar Orthopaedic and Sports Medicine Hospital, Doha, Qatar
c
Department of Rehabilitation Sciences and Physiotherapy Ghent, Ghent University, Ghent, Belgium

A R T I C L E I N F O A B S T R A C T

Keywords: Rehabilitation after injury and reconstruction to the anterior cruciate ligament is thought to require a gradual
AlterG reintroduction of loading, particularly during resumption of running. One strategy to achieve this is via the use
EMG of a reduced-gravity treadmill but it is unknown, if and how muscle activity varies in the reduced gravity
ACL conditions compared to regular treadmill running. Nineteen healthy participants and 18 male patients at the end
Rehabilitation
of their rehabilitation (8 with a bone-patellar-bone graft, 10 with a hamstring graft) participated in this multi-
Return to play
muscle surface electromyography (sEMG) running study. The hamstrings and triceps surae were evaluated
during a 16 km/h running while at 6 dierent relative bodyweight conditions from 50% (half weight-bearing) to
100% (full weight-bearing). Muscle activation was examined individually as well as normalized to a composite
entire activation and considered across the entire gait cycle using Statistical Parametric Mapping. The healthy
participants showed dierences between the 50100% BW and 60100% conditions and in the hamstring graft
group for 60100% and 80100% conditions. No dierences were seen comparing all loading conditions in the
bone-patellar-bone graft group. For the hamstrings, from 70% BW and above, there appear to be no dierence in
activation patterns for any of the groups.
The activation patterns of the hamstrings was essentially the same from 70% indicated bodyweight through to
full weight bearing when running at 16 km/h. Accordingly, when running at this relatively high speed, we do not
expect any adverse eects in terms of altered motor patterns during rehabilitation of these muscles.

1. Introduction
Following Anterior cruciate ligament reconstruction (ACLR), ath-
letes are known to have decits in strength [1], proprioception [2], gait
parameters [3] and muscle function (e.g. quadriceps and hamstrings)
[4]. However, there is no accepted gold standard rehabilitation pro-
tocol for how to best address these decits [5]. Post-operative re-
habilitation is thought to inuence the ultimate outcome [5] with re-
turn to play for athletes post-ACLR being as low as 55% [6]. Long-term
consequences such as increased incidence of arthritis are concerning
[7] and while a graded return to sport is considered a cornerstone of
rehabilitation programs following ACLR [8], there has been little re-
search to objectively quantify this progression back to running.
Gradual re-introduction of running at full bodyweight post ACLR
makes intuitive sense in an attempt to restore gait kinematics, facilitate
muscle function, and decrease the incidence of anterior knee pain [9].
However, while it is known that Increasing running speed during over-
ground and treadmill running [10] results in increased muscle activity
by the athlete [11], little is known about the eects of altering body-
weight while running post ACLR.
One approach to reintroduce running loads gradually is via a weight
supporting treadmill. More specically, positive pressure, or anti-
gravity treadmills allow creation of a controlled weight-bearing en-
vironment to train earlier in the recovery period [12]. Participants wear
snug-tting neoprene shorts sealed into an enclosed treadmill which
has a load cell in the deck, under the belt. After calibration, the device
can deliver an appropriate air pressure such that the participant is
pushed away from the deck thereby reducing their eective body
weight (BW) by the desired amount. The amount of reduction in BW
commonly used can range from no reduction (100% BW) all the way
down to 20% of BW. These devices have been used to study the eect of
weight support on metabolic demand [13], joint forces [14] and plantar
loading or vertical ground reaction force [15].
While reduced impact loading may allow for earlier resumption of
running, it has been suggested that running at reduced BW may be
associated with alterations in motor patterns [16] which, if adopted

Corresponding author.
E-mail address: C.Hansen@neurologie.uni-kiel.de (C. Hansen).

http://dx.doi.org/10.1016/j.gaitpost.2017.07.120
Received 16 March 2017; Received in revised form 25 July 2017; Accepted 28 July 2017
0966-6362/ 2017 Elsevier B.V. All rights reserved.
C. Hansen et al.
permanently, could be ultimately deleterious to an athletes health once
they resume normal running. Of particular interest in the sporting
context is the possibility of muscle injury associated with running; the
highest burden of which falls to the hamstring muscles [17]. The
muscular work performed is shown to be non-linearly related to run-
ning speed with an increasingly higher load as speed increases [18].
Some research has focused on the EMG motor patterns during walking
in a reduced gravity setting in healthy participants [19] and patients
[20], but to our knowledge there are no studies focused on running in
an ACLR patient cohort.
ACLR involves replacing the torn ACL and two common techniques
used are the Bone-patellar tendon-bone (BTB) and the hamstring (HS)
autograft ACL reconstruction [21]. Patients with an ACLR show altered
movement patterns that may be related to the trauma caused by the
surgery (to the hamstrings and quadriceps muscles respectively) and
therefore diverse antalgic responses protecting the exors/extensors are
common. Since hamstring and calf muscle activation inuence knee
loading, its of interest to examine these activation patterns which may
eect risk for ACL re-injury.
Therefore, the aim of this study was to examine how muscle acti-
vation is altered during rehabilitation using dierent bodyweight in
healthy male participants and two patient groups with BTB and HS graft
after they have passed a criteria based rehabilitation program [5]. This
paper focuses on muscle activation of the previously injured and non-
injured leg to examine how each ACLR technique may aect the
function of these muscles which are synergists of the ACL.
2. Methods
Control group: Nineteen injury-free male participants (age
35.4 7.8 years, weight 77.6 8.4 kg, height 179.1 5.6 cm).
Patients: Eighteen male participants who had undergone a recent ACL
reconstruction and had completed all clinical criteria which allowed
resumption of running. The patient group was subsequently divided
into patients that underwent a HS graft (n = 10) (age 26 3.84 years,
weight 74.16 7.19 kg, and height 176.89 5.6 cm) or a BTB re-
construction (n = 8) (age 27 7.69 years, weight 80.40 9.44 kg,
and height: 178.49 7.29 cm). At the time of testing the patients
were, on average, 7 months ( 2) post-surgery. Informed consent was
obtained for each participant, and the experiment was conducted with
the approval of the local ethics committee (ASPETAR: F2013000001).
All participants volunteered to take part in this study and both patient
groups were discharged from the rehabilitation facility and declared as
ready to return to on-eld sports specic activity.
EMG activity was recorded by means of surface electrodes from 6
muscles on both legs of the participant (i.e. 12 EMG channels). These
included the plantar exors soleus medialis (SM), soleus lateralis (SL),
medial gastrocnemius (MG), lateral gastrocnemius (LG) and the ham-
string muscles: medial and lateral hamstrings. The electrodes were
placed by palpating the muscle bellies and orienting the electrodes
along the main direction of the bers [22].
The EMG sensors also contained 3D accelerometers, these inertial
signals from the sensor placed on the shin in order to dene the gait
cycle during running [23]. All data was recorded at 2000 Hz using a
Delsys Trigno Wireless System (Boston, MA), and ltered (high pass
lter with 5 Hz cut o, low pass with 500 Hz cuto). Heel strike was
identied by minima in the shank vertical acceleration.
The running data was continuously monitored and each step was
extracted and normalized from heel strike to heel strike leading to 101
separate time points (0100%) (Fig. 1).
Six running trials were performed on an Alter G treadmill at a
running speed of 16 km/h with varying BW support ranging from 50%
BW up to 100% in 10% increments. The 100% BW running condition is
the reference condition comparable to regular treadmill running
without any support. To assess relative muscle activation across the
dierent loading conditions, the EMG signal was normalized to its
155
Gait & Posture 58 (2017) 154158
respective maximal contraction value obtained during the 100% BW
running trials for each participant.
Trial conditions were presented in a random order for each parti-
cipant. In each condition, after an initial familiarization and once
participants reported feeling comfortable in the treadmill, measure-
ments started. At this time the participants ran for a minimum of 6
(maximum 30) further strides. Data from all of these strides for each
condition, for each participant were individually averaged for sub-
sequent analysis.
Comparative EMG analysis was conducted for the entire time-de-
pendent EMG signal using Statistical Parametric Mapping (SPM, version
M.03) in Matlab (R2014a, The Mathworks Inc, Natick, MA). More ty-
pical statistical approaches are shown to be problematic when there are
many comparisons made. SPM [2427] is a method that allows analysis
of time-series without reducing the dimensionality to only specic
scalars (e.g. peak amplitude swing/stance phase). This approach was
adopted to allow comparison of each 1% interval of the dierent run-
ning conditions without suering from Type I errors (due to multiple
comparisons). SPM and embedded vector-eld analysis (multiple time-
dependent muscle signals clustered within one vector) provides better
control of type I and type II statistical error where multiple comparisons
are conducted [28]. Clusters were formed based on the similar time-
dependency and synergistic activation characteristics and to form time-
dependent muscle activity vectors. We clustered the EMG-series of the
SM and SL, the GM and GL and the medial and lateral hamstrings
forming a Soleus, a Gastrocnemius and a Hamstring cluster, respec-
tively.
The mean normalized muscle activity amplitude (per cluster and
running cycle) and corresponding standard deviations at each of the
101 data frames is used for the statistical comparison of the entire time-
dependent EMG signal (Fig. 1). To identify intra-group dierences (left
vs right leg, injured vs uninjured leg) of the multi-muscle EMG series
paired Hotellings T2 statistics were performed. Following vector eld
analyses, post hoc tests were conducted on each vector component se-
parately (paired T-statistic). The signicance level was set to p < 0.05.
3. Results
The SPM vector eld analysis of the cluster with the highest con-
dition i.e. 50100%; 60%100%; 70100%; 80%100% and 90100%
are presented below (Fig. 2 and for more detailed presentation see the
Supplementary data le).
Control: Dierences were seen for the control group comparing the
50100% BW and 60100% conditions but no dierences were seen
above especially not in the Hamstring cluster.
BTB: No dierences were seen comparing all loading conditions in
the BTB group.
HS: Clear dierences were seen for the hamstring autograft group
comparing 50% up to 80% BW.
Above 80%, there appear to be no dierence in activation patterns
for any of the groups and clusters.
3.1. Increasing load dependencies
When comparing the cluster within the next higher condition i.e.
5060%; 60%70%; 7080%; 80%90% and 90100% the SPM vector
eld analysis of the three clusters found no signicant dierences for all
three cohorts between the six running conditions (Fig. 2).
4. Discussion
The aim of this study was to examine if relative loading inuences
muscle activation patterns during running in control and ACLR parti-
cipants. When comparing the muscle activity of the triceps surae and
hamstrings in this study, signicantly dierent levels of normalized
EMG activity were observed between the dierent loading conditions
C. Hansen et al. Gait & Posture 58 (2017) 154158

Fig. 1. (a) The muscle activity pattern of the three cohorts is shown throughout the six weight bearing conditions (50100% BW). Red lines and shading indicate the non-injured/left leg
while green represents the injured/right leg. (b) shows an exemplary statistical comparison with the eect size computation between the clustered muscle activities for the 50% loading
condition with the 100% condition. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

Fig. 2. SPM vector eld analysis of the Soleus, Gastrocnemius and Hamstring clusters found clear dierences between loading condition for each cohort. The activation patterns of the
three muscle clusters showed no dierences with decreased body weight support at constant running speed for the BTB, while no dierences were found from 70% onwards for the control
group and no dierences from the 90% BW condition for the hamstring autograft cohort. The black dashed border serve for visual representation where no dierences were found and the
red dashed line indicates the critical random eld theory threshold. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this
article.)

156
C. Hansen et al.
for healthy controls. These results are in accordance with the ndings of
Liebenberg et al. [29] who used the average and root mean square as
parameters and showed that the BW support in percentage does not
correspond to the reduction in muscle activity (bicep femoris, rectus
femoris, tibialis anterior, and gastrocnemius). These ndings contrast
with the work of Hunter et al. [19] who collected EMG data on patients
and a control group during a 16 km/h run on an AlterG treadmill at
50%, 60%, 70%, 80% and 100% indicated body weight. Hunter [19]
documented a signicant increase of EMG amplitude for the gastro-
cnemius, tibialis anterior, peroneus longus, vastus medialis, vastus la-
teralis and the hip abductors with decreasing BW support. Interestingly,
those ndings could not be reproduced entirely for the BTB group and
the HS group.
These results suggest that whilst the patients were ready to be dis-
charged from rehabilitation and declared as ready to return to play,
muscle activation had potentially not yet normalized given the dier-
ences observed in the control group. This information may allow a re-
nement of return to play criteria and further reduce re-injury rates [5].
It has been shown that vector eld SPM and scalar extraction results
yielded dierent results and even in compatible statistical conclusions
and as such previous ndings about positive pressure treadmills could
be biased based on the chosen method and the interpretation of the
results.
Running at 70% BW shows similar muscle activation patterns as
100% BW running while reducing the load on the joint. We therefore
propose that running at 70% BW allows a sensible tradeo for re-
habilitation as the muscle activity is similar to normal running while
the loads are still reduced. This is an important nding as it shows that
positive-pressure treadmills can be used early on in the rehabilitation
process to provide exposure to higher speed running thereby helping
regain function and consequently improve the return to play process
without overloading the patients.
Focusing on the EMG amplitude only and reducing the muscle ac-
tivity time series to scalars provides an incomplete picture of the un-
derstanding of the motor activation pattern. In our study, SPM vector
eld analyses was used that allows comparison of the entire time-de-
pendent EMG signal. This analysis is based on Random Field Theory
and has been introduced recently in biomechanical data analysis [26].
Initially this method was used to analyze functional brain images but
has been commonly used in biomechanics and especially in EMG ana-
lysis [25]. The advantage of SPM is that it considers vector covariance
when computing the test statistic and eld smoothness and size when
computing the critical test statistic threshold. The SPM analysis mini-
mises Type I and Type II errors not only because of the aforementioned
facts but also it considers random eld behavior when computing p
values while using the entire measurement domain and thus main-
taining a constant error of [27].
This is the rst study focussing on the inuence of weight-bearing
running conditions on the muscle activation patterns using three in-
dependent cohorts while applying a new statistical approach allowing
to compare EMG time-series. Peak muscle activation or similar time
unrelated scalars do not give a complete insight about how the muscle
activation changes but only shows that body weight support decreased
muscle activity [16].
5. Limitations
The normalization of EMG signals is generally a limitation and has
been extensively discussed over the years [30]. The choice to normalize
the EMG data based on the maximal amplitude encountered during the
running trials does not allow us to compare between groups. However
the comparison between conditions (50, 60, 70, 80, 90 & 100% BW)
clearly showed that the muscle activity does change with weight-
bearing support for the controls but not for the BTB group. In this study
we have only considered one running speed 16 km/h that was
thought to reect a reasonable clinical eort. These results may not be
157
Gait & Posture 58 (2017) 154158
the same for dierent speeds, or for walking and accordingly should not
be extrapolated. It cannot be assumed that the results presented here
hold for all muscles involved in locomotion. Future research could also
target the quadriceps muscles and analyze the EMG activity for each
muscle individually to avoid clusters. Previous research has docu-
mented dierent activation patterns during reduced loading for muscle
groups not examined here [19] and an interesting way to evaluate
motor patterns could be by using principal component analysis or non-
negative matrix factorization. Finally, additional kinematic/kinetic
analysis may have given more insight in the movement patterns and the
underlying muscle activity.
6. Conclusions
In control participants dierences in activation patterns were seen
when comparing the 50% BW and 60% BW conditions to each of the
higher BW conditions. In BTB ACLR participants, there were no dif-
ferences in activation patterns seen for any comparison. In HS ACLR
participants, dierences were observed comparing the 60% BW con-
dition to 100% for the gastrocnemius and hamstring cluster of the in-
jured leg. These data suggest that individual assessment of activation
patterns is required to tailor a targeted rehabilitation program.
Conict of interest
The authors declare no conict of interest.
Acknowledgements
The authors express their appreciation to the patients and partici-
pants for participating this study.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.gaitpost.2017.07.120.
References
[1] A. Natri, M. Jrvinen, K. Latvala, P. Kannus, Isokinetic muscle performance after
anterior cruciate ligament surgery: long-term results and outcome predicting factors
after primary surgery and late-phase reconstruction, Int. J. Sports Med. 17 (1996)
223228, http://dx.doi.org/10.1055/s-2007-972836.
[2] M. Ochi, J. Iwasa, Y. Uchio, N. Adachi, Y. Sumen, The regeneration of sensory
neurones in the reconstruction of the anterior cruciate ligament, J. Bone Joint Surg.
Br. 81 (1999) 902906.
[3] P. DeVita, T. Hortobagyi, J. Barrier, Gait biomechanics are not normal after anterior
cruciate ligament reconstruction and accelerated rehabilitation, Med. Sci. Sports
Exerc. 30 (1998) 14811488.
[4] M. Lewek, K. Rudolph, M. Axe, L. Snyder-Mackler, The eect of insucient
quadriceps strength on gait after anterior cruciate ligament reconstruction, Clin.
Biomech. Bristol Avon. 17 (2002) 5663.
[5] P. Kyritsis, R. Bahr, P. Landreau, R. Miladi, E. Witvrouw, Likelihood of ACL graft
rupture: not meeting six clinical discharge criteria before return to sport is asso-
ciated with a four times greater risk of rupture, Br. J. Sports Med. 50 (2016)
946951, http://dx.doi.org/10.1136/bjsports-2015-095908.
[6] C.L. Ardern, N.F. Taylor, J.A. Feller, K.E. Webster, Fifty-ve per cent return to
competitive sport following anterior cruciate ligament reconstruction surgery: an
updated systematic review and meta-analysis including aspects of physical func-
tioning and contextual factors, Br. J. Sports Med. 48 (2014) 15431552, http://dx.
doi.org/10.1136/bjsports-2013-093398.
[7] S.R. Filbay, I.N. Ackerman, T.G. Russell, K.M. Crossley, Return to sport matters-
longer-term quality of life after ACL reconstruction in people with knee diculties,
Scand. J. Med. Sci. Sports (2016), http://dx.doi.org/10.1111/sms.12698.
[8] P. Glasgow, N. Phillips, C. Bleakley, Optimal loading: key variables and mechan-
isms, Br. J. Sports Med. 49 (2015) 278279, http://dx.doi.org/10.1136/bjsports-
2014-094443.
[9] Y. Kaplan, Y. Barak, E. Palmonovich, M. Nyska, E. Witvrouw, Referent body weight
values in over ground walking, over ground jogging, treadmill jogging, and ellip-
tical exercise, Gait Posture 39 (2014) 558562, http://dx.doi.org/10.1016/j.
gaitpost.2013.09.004.
[10] T.W. Kernozek, K.A. Zimmer, Reliability and running speed eects of in-shoe
loading measurements during slow treadmill running, Foot Ankle Int. 21 (2000)
749752.
C. Hansen et al.
[11] C. Hansen, E. Einarson, A. Thomson, R. Whiteley, Peak medial (but not lateral)
hamstring activity is signicantly lower during stance phase of running. An EMG
investigation using a reduced gravity treadmill, Gait Posture 57 (2017) 710,
http://dx.doi.org/10.1016/j.gaitpost.2017.05.012.
[12] B. Gojanovic, R. Shultz, F. Feihl, G. Matheson, Overspeed HIIT in lower-body po-
sitive pressure treadmill improves running performance, Med. Sci. Sports Exerc. 47
(2015) 25712578, http://dx.doi.org/10.1249/MSS.0000000000000707.
[13] D.K.P. McNeill, H.D. de Heer, C.P. Williams, J.R. Coast, Metabolic accommodation
to running on a body weight-supported treadmill, Eur. J. Appl. Physiol. 115 (2015)
905910, http://dx.doi.org/10.1007/s00421-014-3071-y.
[14] S. Patil, N. Steklov, W.D. Bugbee, T. Goldberg, C.W. Colwell, D.D. DLima, Anti-
gravity treadmills are eective in reducing knee forces, J. Orthop. Res. O. Publ.
Orthop. Res. Soc. 31 (2013) 672679, http://dx.doi.org/10.1002/jor.22272.
[15] A. Thomson, E. Einarsson, E. Witvrouw, R. Whiteley, Running speed increases
plantar load more than per cent body weight on an AlterG treadmill, J. Sports Sci.
(2016) 16, http://dx.doi.org/10.1080/02640414.2016.1163401.
[16] P. Sainton, C. Nicol, J. Cabri, J. Barthelemy-Montfort, E. Berton, P. Chavet,
Inuence of short-term unweighing and reloading on running kinetics and muscle
activity, Eur. J. Appl. Physiol. 115 (2015) 11351145, http://dx.doi.org/10.1007/
s00421-014-3095-3.
[17] J. Ekstrand, M. Hagglund, M. Walden, Injury incidence and injury patterns in
professional football: the UEFA injury study, Br. J. Sports Med. 45 (2011) 553558,
http://dx.doi.org/10.1136/bjsm.2009.060582.
[18] A.G. Schache, P.D. Blanch, T.W. Dorn, N.A.T. Brown, D. Rosemond, M.G. Pandy,
Eect of running speed on lower limb joint kinetics, Med. Sci. Sports Exerc. 43
(2011) 12601271, http://dx.doi.org/10.1249/MSS.0b013e3182084929.
[19] I. Hunter, M.K. Seeley, J.T. Hopkins, C. Carr, J.J. Franson, EMG activity during
positive-pressure treadmill running, J. Electromyogr. Kinesiol. O. J. Int. Soc.
Electrophysiol. Kinesiol. 24 (2014) 348352, http://dx.doi.org/10.1016/j.jelekin.
2014.01.009.
[20] M.P. Berthelsen, E. Husu, S.B. Christensen, K.P. Prahm, J. Vissing, B.R. Jensen, Anti-
gravity training improves walking capacity and postural balance in patients with
muscular dystrophy, Neuromuscul. Disord. 24 (2014) 492498, http://dx.doi.org/
10.1016/j.nmd.2014.03.001.
[21] J. Kautzner, P. Kos, M. Hanus, T. Trc, V. Havlas, A comparison of ACL re-
construction using patellar tendon versus hamstring autograft in female patients: a
prospective randomised study, Int. Orthop. 39 (2015) 125130, http://dx.doi.org/
158
Gait & Posture 58 (2017) 154158
10.1007/s00264-014-2495-7.
[22] H.J. Hermens, B. Freriks, C. Disselhorst-Klug, G. Rau, Development of re-
commendations for SEMG sensors and sensor placement procedures, J.
Electromyogr. Kinesiol. O. J. Int. Soc. Electrophysiol. Kinesiol. 10 (2000)
361374.
[23] F. Sylos-Labini, V. La Scaleia, A. dAvella, I. Pisotta, F. Tamburella, G. Scivoletto,
M. Molinari, S. Wang, L. Wang, E. van Asseldonk, H. van der Kooij, T. Hoellinger,
G. Cheron, F. Thorsteinsson, M. Ilzkovitz, J. Gancet, R. Haue, F. Zanov,
F. Lacquaniti, Y.P. Ivanenko, EMG patterns during assisted walking in the exoske-
leton, Front. Hum. Neurosci. 8 (2014), http://dx.doi.org/10.3389/fnhum.2014.
00423.
[24] Statistical Parametric Mapping: The Analysis of Functional Brain Images, Reprinted
(twice), in: K.J. Friston (Ed.), Elsevier, Acad. Press, Amsterdam, 2008.
[25] J. Schuermans, L. Danneels, D. Van Tiggelen, T. Palmans, E. Witvrouw, Proximal
neuromuscular control protects against hamstring injuries in male soccer players: a
prospective study with electromyography time-series analysis during maximal
sprinting, Am. J. Sports Med. 45 (2017) 13151325, http://dx.doi.org/10.1177/
0363546516687750.
[26] T.C. Pataky, Generalized n-dimensional biomechanical eld analysis using statis-
tical parametric mapping, J. Biomech. 43 (2010) 19761982, http://dx.doi.org/10.
1016/j.jbiomech.2010.03.008.
[27] T.C. Pataky, M.A. Robinson, J. Vanrenterghem, Vector eld statistical analysis of
kinematic and force trajectories, J. Biomech. 46 (2013) 23942401, http://dx.doi.
org/10.1016/j.jbiomech.2013.07.031.
[28] T.C. Pataky, Assessing the signicance of pedobarographic signals using random
eld theory, J. Biomech. 41 (2008) 24652473, http://dx.doi.org/10.1016/j.
jbiomech.2008.05.010.
[29] J. Liebenberg, J. Scharf, D. Forrest, J.S. Dufek, K. Masumoto, J.A. Mercer,
Determination of muscle activity during running at reduced body weight, J. Sports
Sci. 29 (2011) 207214, http://dx.doi.org/10.1080/02640414.2010.534806.
[30] M. Halaki, K. Gi, Normalization of EMG signals: to normalize or not to normalize
and what to normalize to? in: G.R. Naik (Ed.), Comput. Intell. Electromyogr. Anal.
Perspect. Curr. Appl. Future Chall. InTech, 2012http://www.intechopen.com/
books/computational-intelligence-in-electromyography-analysis-a-perspective-on-
current-applications-and-future-challenges/normalization-of-emg-signals-to-
normalize-or-not-to-normalize-and-what-to-normalize-to- (Accessed 23 June
2016).