Medical Engineering & Physics 21 (1999) 247–257

www.elsevier.com/locate/medengphy

Normalisation of EMG amplitude: an evaluation and comparison of

old and new methods
a,* b
Adrian Burden , Roger Bartlett
a
Chelsea School Research Centre, University of Brighton, Chelsea School, Hillbrow, Denton Road, Eastbourne BN20 7SR, UK
b
Sport Science Research Institute, Sheffield Hallam University, Sheffield S10 2BP, UK

Received 15 April 1998; received in revised form 22 June 1999; accepted 6 July 1999

Abstract

The purpose of this study was to evaluate and compare four different methods of normalising the amplitude of electromyograms
(EMGs), from the biceps brachii. Five males performed isotonic contractions of the elbow flexors with an external force of 50 N,
100 N, 150 N and 200 N. These were followed by a single isometric maximal voluntary contraction (MVC) and ten isokinetic
MVCs at 0.35 rad s⫺1 intervals between 0.35 rad s⫺1 and 3.50 rad s⫺1. The processed EMGs recorded from the isotonic contractions
were normalised by expressing them as a percentage of: (i) the mean (Dynamic Mean Method) and (ii) the peak EMG from the
same contraction (Dynamic Peak Method), (iii) the EMG from the isometric MVC (Isometric MVC Method), and (iv) the EMG
from an isokinetic MVC at the same elbow angle and angular velocity (Isokinetic MVC Method). The root mean square difference
(RMSD) between the outputs of the Isokinetic MVC and Dynamic Mean methods was significantly greater (P⬍0.05) than between
the Isokinetic MVC method and the Dynamic Peak and the Isometric MVC methods. The small (10%) difference between the
Isokinetic MVC and the Isometric MVC Methods was a consequence, firstly, of the lack of difference in EMG recorded from the
isometric and isokinetic MVCs and, secondly, the consistency in EMG over the range of motion and at different angular velocities
of isokinetic MVC. We conclude that only the Isometric and Isokinetic MVC methods should be used to normalise the amplitude
of EMGs from the biceps brachii.  1999 IPEM. Published by Elsevier Science Ltd. All rights reserved.

Keywords: Electromyography; Normalisation; Biceps brachii; Isotonic contraction; Isometric contraction; Isokinetic contraction

1. Introduction activity of a single muscle from the same individual

The amplitude and frequency characteristics of the
raw electromyogram (EMG) detected using surface elec-
trodes has been shown to be sensitive to many intrinsic
and extrinsic factors [1]. Muscle fibre type, diameter,
depth and location with respect to the electrodes, and the
amount of tissue between the muscle and the electrode
are some of the intrinsic factors which cannot be con-
trolled. Extrinsic factors, which can be influenced by the
experimenter, include the orientation, location, area and
shape of the electrodes, and the distance between them.
Thus, the amplitude of the temporally processed EMG
can only be used to assess short term changes in the
* Corresponding author Tel.: +44-1273-643715; fax: +44-1273-
643704.
E-mail address: a.burden@brighton.ac.uk (A. Burden)
when the electrode set up has not been altered [2].
To allow comparison of activity between different
muscles, across time, and between individuals, the EMG
should be normalised [1–5], i.e. expressed in relation to
a reference value obtained during standardised and
reproducible conditions [2]. Early investigations of
dynamic tasks, including walking [6,7], have used the
EMG from an isometric maximum voluntary contraction
(MVC) as the normalisation reference value. However,
it is generally recognised that the EMG from an iso-
metric MVC is less reliable than the signal obtained from
an isometric submaximal contraction [8], and that it
might not represent the maximum activation capacity of
the muscle [9]. This has led to the evaluation and use
of other reference values; in addition, some authors
[5,10] have expressed alternative aims for the normalis-
ation of the EMG.
Yang and Winter [5] compared four different normal-
isation reference values to see which would result in the

1350-4533/99/$ - see front matter.  1999 IPEM. Published by Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 0 - 4 5 3 3 ( 9 9 ) 0 0 0 5 4 - 5
248 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
greatest reduction in inter-subject variability during
walking. The use of either the mean or the peak linear
envelope from the ensemble average of at least six
strides reduced the inter-subject coefficient of variation
in relation to the un-normalised data in all five lower
limb muscles analysed. In comparison, the inter-subject
coefficient of variation was generally increased by using
either 50% of the isometric MVC or the mean EMG
per unit of isometric moment as the reference values.
A reduced inter-subject coefficient of variation was also
demonstrated for the biceps brachii during isotonic
elbow flexions and extensions [11] and for the gastro-
cnemius during a balancing task [3] by using the peak
or mean ensemble value in comparison to the EMG from
an isometric submaximal [11] or MVC [3,11]. Although
the peak and mean ensemble methods are the only feas-
ible ways of normalising EMGs from patients with neur-
ologic disorders [5], they tend to produce a normal EMG
template for a particular task and, therefore, may remove
the true biological variation within a group [3,11]. How-
ever, normalisation of EMGs using an isometric MVC
has been shown to increase the intra-class correlation
coefficient between trials, in comparison to the peak and
mean dynamic values, thereby improving the reproduci-
bility of the data [3]. Although the isometric MVC
method is the only one that aims to reveal the percentage
of the maximum activation capacity of the muscle
required to perform a specific task [5], all other methods
mentioned above generally reflect changes in the un-nor-
malised data as a consequence of changes in load and
velocity of movement [11].
Early studies have shown that the EMG from the
elbow flexors is largely unaffected by joint angle [12],
type of muscle action (i.e. concentric or eccentric), or
angular velocity [13] during isokinetic MVCs. Con-
versely, it has been well established that the EMG
recorded during isokinetic MVCs of knee extensor
muscles is largest in the mid range of motion [14–16],
is greater for concentric in comparison to eccentric con-
tractions [15,17–19], and increases with increasing angu-
lar velocity during concentric contractions [15,17–19].
The EMG from an isometric MVC may not, therefore,
represent the maximum activation capacity of the muscle
either at lengths other than those at which the MVC was
performed, or under non-isometric conditions. Evidence
of this was partially provided by Mirka [4], who normal-
ised trunk muscle EMGs from dynamic flexions using
isometric MVCs, which were performed at the same
trunk angle and at an arbitrary trunk angle. Mean differ-
ences of between 15 and 50% were reported between
the two methods, depending on the muscle involved.
Furthermore, Kellis and Baltzopoulos reported signifi-
cant differences between EMGs from the hamstrings,
normalised using either a single angle isometric MVC
or isokinetic MVCs performed at the same angular velo-
city, a similar angle and with same type of muscle action
[20]. Research into the factors that affect the magnitude
of the EMG recorded from the elbow flexors during iso-
kinetic MVCs is, therefore, sparse and disagrees with the
more recent and abundant research undertaken using the
knee extensor muscle group.
The overall aim of this study was to investigate the
relationship between EMG and joint angle, angular velo-
city and type of muscle action in the biceps brachii, and
to use this muscle to evaluate and compare a number
of EMG amplitude normalisation methods. The specific
objectives were:
1. To determine, experimentally, the effect that joint
angle, angular velocity, and type of muscle action
has on EMGs from isokinetic MVCs of the biceps
brachii. Previous research which addressed a similar
objective [12,13] was undertaken over 25 years ago
and used an isokinetic dynamometer which claimed
to maintain constant linear velocity of muscle con-
traction rather than angular velocity of forearm
rotation. The biceps was, therefore, selected in order
to corroborate, or otherwise, the early findings of
Komi and colleagues [12,13]. This would also help
to ascertain whether the non-uniform relationships
discovered between EMG and joint angle [14–16],
angular velocity and type of muscle action [15,17–
19], for the knee extensors, were muscle dependant.
2. To compare the magnitude of EMGs normalised by
expressing them as a percentage of the EMGs
recorded from isokinetic MVCs at the identical
angle and angular velocity, and from the same type
of muscle action, with those normalised using other
normalisation methods that have previously been
evaluated and compared.
3. To compare the inter-subject variability of EMGs
normalised by the same methods as expressed in the
second objective.
4. To determine how the magnitude of force set for a
submaximal isotonic task, and type of muscle action,
affected the comparisons between normalisation
methods.
2. Methods
2.1. Subjects
Five males ((mean±SD) age 23.3±3.8 yrs; height
1.78±0.06 m; mass 84.2±6.2 kg) took part in the investi-
gation after reading and signing an informed consent
document. All volunteers had recently taken part in
experiments which had required them to perform iso-
metric and isokinetic MVCs of the elbow flexor muscles
using identical equipment to that used in this study.
 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
2.2. Procedures
Submaximal isotonic, and isometric and isokinetic
MVCs of the elbow flexor muscle group were performed
using a KIN COM 500H dynamometer (Chattanooga
Group Inc., Hixson, Tennessee). The KIN COM and
associated software was used to sample force, measured
by the dynamometer’s load cell situated laterally to the
distal end of the forearm, and angular position and angu-
lar velocity of the forearm. The sampling frequency was
100 Hz during each contraction. An MT8 Radio Tel-
emetry System (MIE Medical Research Ltd., Leeds,
Yorkshire) connected to the KIN COM, was used to rec-
ord EMGs from the biceps brachii, which were
synchronised with the angular position and angular velo-
city data. The MT8 System included a differential pre-
amplifier (gain ×1000, input impedance ⬎10 M⍀, com-
mon mode rejection ratio ⬎100 dB) connected to the
electrodes at the skin, which was plugged into a trans-
mitter worn around the subject’s waist. The transmitter
was fitted with an anti-aliasing second order Chebyshev
low-pass filter set at a cut-off frequency of 1000 Hz.
High-pass filtering, set by the limits of the pre-ampli-
fiers, was typically about 5 Hz. The filtered signal was
transmitted to the nearby receiver, connected to a PC
via a 12 bit analog-to-digital expansion board (Amplicon
PC26AT) which sampled the signal at 2000 Hz.
All contractions were performed while the subject was
seated and stabilised, using chest and waist straps, on
the KIN COM with the upper arm flexed to a horizontal
position. The forearm was supinated and the lateral
epicondyle of the humerus was aligned with the axis of
rotation of the dynamometer.
As recommended by Clarys and Cabri [21], bipolar
surface electrodes were placed over the visual midpoint
of the contracted belly of the biceps brachii of the domi-
nant arm with an angle of flexion of 90° (full
extension=0°). Electrodes were circular domes
(area=113 mm 2) made from a tin-lead alloy, positioned
with a centre-to-centre inter-electrode distance of 2 cm
along a line approximately parallel to the direction of
the underlying muscle fibres. Before electrode place-
ment, the skin underlying the electrode sites was shaved
and cleaned with an alcohol wipe. To reduce the skin
resistance to below 10 k⍀ the skin was lightly scratched
with a sterile lancet, as recommended by Okamoto et al.
[22], and the domes of the electrodes were filled with
Parker Signa Electrode Gel.
Initially, a series of four submaximal isotonic
(concentric followed by eccentric) contractions of the
elbow flexors of the dominant arm were performed over
a range of motion of 80° (20° to 100° of elbow flexion)
with the KIN COM set to provide constant resistances
of 50 N, 100 N, 150 N and, finally, a near maximal 200
N. Secondly, following a 5 minute rest, an isometric
MVC of the elbow flexors was performed at 90° of
249
elbow flexion, reaching maximal force as rapidly as
possible and maintaining it for 3 s. Finally, a series of
isokinetic MVCs (concentric followed by eccentric) of
the elbow flexors were performed at 0.35 rad s⫺1 inter-
vals between 0.35 rad s⫺1 and the highest angular velo-
city reached by the subject during the previous submaxi-
mal isotonic contractions. As the highest angular
velocity recorded during the isotonic trials was never
greater than 3.50 rad s⫺1, the maximum number of isoki-
netic MVCs (including concentric and eccentric) perfor-
med by any subject was ten. The order of angular velo-
cities was randomised and a rest period of five minutes
was given between each pair of concentric and eccentric
contractions to avoid the effects of fatigue. Each isoki-
netic MVC was performed over a 100° range of motion
(10° to 110° of elbow flexion) to ensure minimal acceler-
ation and deceleration of the forearm within the smaller
range of motion that was used during the previous iso-
tonic contractions. Subjects were instructed to pull as
hard and as fast as they could throughout the full range
of motion and were able to view an on-line record of
force on the KIN COM monitor.
2.3. Data processing and normalisation
A synchronised record of angle and angular velocity
of elbow flexion, force and biceps brachii raw EMGs
were collected from each of the isotonic contractions and
from the isometric and isokinetic MVCs. The raw EMGs
from each of these contractions were then full-wave rec-
tified and smoothed using a moving average, with
Microsoft Excel software, throughout the full range of
motion. This moving average “window” had a width of
100 ms and was moved along the rectified EMG every
10 ms. The effect that this processing method (commonly
referred to as the mean absolute value) had on the raw
EMG from a single isotonic contraction, together with
the synchronised angle and angular velocity of elbow
flexion and the external force is shown in Fig. 1.
The processed EMGs from the submaximal isotonic
contractions were normalised by expressing them as the
percentage of the following four processed EMG values:
1. The mean EMG recorded from the same isotonic
trial (Dynamic Mean Method).
2. The peak EMG recorded from the same isotonic trial
(Dynamic Peak Method).
3. The maximum EMG recorded from the isometric
MVC (Isometric MVC Method).
4. The EMG recorded from an isokinetic MVC at the
identical elbow angle and angular velocity
(Isokinetic MVC Method). Isotonic EMGs were
matched with EMGs from isokinetic MVCs of the
same angle and angular velocity for this method
using Microsoft Access software.
It was decided not to include a further method, using the
250 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
Fig. 1. Synchronised record of angle and angular velocity of elbow
flexion and extension, external force and biceps brachii EMG for the
concentric and eccentric phases of a single isotonic contraction.
EMG from a submaximal isometric voluntary contrac-
tion, for several reasons: This method is little better at
reducing inter-subject variability than the Isometric
MVC Method [11], and is worse than the Dynamic Peak
and Mean methods [5,11]. It also does not reveal the
percentage of the maximum activation capacity of the
muscle required to perform a particular task.
2.4. Data analysis
A mean EMG value was obtained for each subject
from the concentric and eccentric phases of each isoki-
netic MVC. Each mean was then expressed as the per-
centage difference from the EMG of the subject’s iso-
metric MVC. These differences were compared across
angular velocities and between concentric and eccentric
contractions using a two within-subjects factor (10×2)
repeated measures analysis of variance (ANOVA).
Comparison between the output of the Isokinetic
MVC Method and the output of the other three normalis-
ation methods was assessed using the root mean square
difference (RMSD) between them. A three within-sub-
jects factor (3×4×2) repeated measures ANOVA was
used to determine if the RMSDs were significantly dif-
ferent between the three comparisons, for each magni-
tude of isotonic force and type of muscle action.
Calculation of Z scores for skewness and kurtosis
revealed that none of the data exhibited significant skew-
ness or kurtosis (P⬎0.05), supporting the assumption of
normal distribution. Violations of the assumption of
sphericity were corrected for using the Huynh–Feldt
Adjustment. All statistical tests were performed using
SPSSx software.
The inter-subject variability of the EMGs from each
normalisation method was calculated, for all magnitudes
of force and both types of muscle action, using the coef-
ficient of variation.
3. Results
Examination of the EMGs for each isokinetic MVC
did not reveal any distinct patterns either within or
between angular velocities for any of the subjects. The
results from one representative subject, shown in Figs.
2 and 3, illustrate that, for most of the angular velocities
and muscle actions, the EMG was relatively constant
over the range of motion.
The effect that angular velocity and type of muscle
action had on the EMGs from the biceps brachii is
presented in Fig. 4. The concentric EMGs differed from
the isometric value by between 2±10% (mean±SD) and
⫺9±14% at 0.35 rad s⫺1 and 0.70 rad s⫺1 respectively.
A gradual decline was observed in the eccentric values
as the angular velocity rose, but the variation was not
statistically significant (F3,13=1.08, P⬎0.05). Similarly,
no significant difference in EMG was discovered
between concentric and eccentric muscle actions
(F1,4=0.44, P⬎0.05), and no significant interaction
existed between angular velocity and muscle action
(F2,7=2.40, P⬎0.05).
The outputs from each of the four different normalis-
ation methods from the concentric and eccentric phases
of the isotonic contractions are shown, for each magni-
tude of force, in Fig. 5. The Dynamic Mean Method pro-
duced the highest normalised EMGs (71±12 to 130±22%
of mean isotonic EMG) because of the smaller denomi-
nator used in the normalisation equation. The outputs
from this and the Dynamic Peak Method were both able
to reflect the greater muscle activity required to perform
 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257 251

Fig. 2. EMG in relation to angle of elbow flexion for the concentric phase of all isokinetic MVCs for one subject. EMG appears more smoothed
at higher angular velocities due to less data points being available for presentation.

the concentric phase, in relation to the eccentric phase,
of each isotonic trial. However, as a consequence of
using a different denominator for each trial, these
methods were unable to mirror the increase in muscle
activity observed in the unnormalised EMGs as the
external force was increased. In comparison, the Iso-
metric MVC Method and Isokinetic MVC Method were
able to mirror the changes seen in the unnormalised
EMGs both between concentric and eccentric muscle
actions and as the external force increased.
Fig. 6 shows the RMSD between the same EMGs nor-
malised using the Isokinetic MVC Method and the other
three methods, for both types of muscle action and at
each magnitude of force. A statistically significant differ-
ence in RMSD between the three comparisons
(F1,5=19.19, P⬍0.05) was observed regardless of force
or type of muscle action. Post hoc comparisons, using
the Games and Howell [23] procedure, revealed that the
comparison with the Dynamic Mean Method resulted in
a significantly (P⬍0.05) greater RMSD (38±23%) than
the comparisons with the Dynamic Peak (17±18%) and
the Isometric MVC (10±8%) methods. The RMSD was
also significantly greater (F1,4=13.59, P⬍0.05) for con-
centric rather the eccentric contractions, irrespective of
force magnitude or comparative normalisation method.
In addition, a statistically significant difference in RMSD
was observed between force magnitudes (F1,6=8.28,
P⬍0.05) regardless of normalisation methods compared
252 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257

Fig. 3. EMG in relation to angle of elbow flexion for the eccentric phase of all isokinetic MVCs for one subject. EMG appears more smoothed
at higher angular velocities due to less data points being available for presentation.

or the type of muscle action. However, a significant
interaction, clearly shown in Fig. 6, existed between the
magnitude of force and the methods that were being
compared (F4,18=29.86, P⬍0.05). In the comparisons
involving the Isokinetic MVC Method with both the
Dynamic Peak and Isometric MVC methods, the RMSD
rose as the force was increased. In contrast, for the com-
parison between the Isokinetic MVC and Dynamic Mean
methods, the RMSD decreased as the force was
increased.
The inter-subject coefficients of variation for the out-
put of the four normalisation methods are shown in
Table 1, for each magnitude of force and for both types
of muscle action. All normalisation methods reduced
inter-subject variability in relation to the unnormalised
EMGs. Over most conditions of force and muscle action,
the Dynamic Mean Method produced the lowest coef-
ficient of variation. Greater inter-subject variability
resulted from the use of the other normalisation methods,
in particular the Isometric MVC Method.
4. Discussion
4.1. EMG-angle and EMG-angular velocity
relationship
The finding that the electrical activity in the biceps
brachii was independent of the angle of elbow flexion
 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257 253

Fig. 4. Mean (±SD) EMG, expressed as a percentage of the EMG from an isometric MVC (%MVC), from the concentric and eccentric phases
of isokinetic MVCs at different angular velocities.

Fig. 5. Mean (+SD) output from the four normalisation methods for concentric and eccentric phases of isotonic contractions at four different forces.

during most of the concentric and eccentric isokinetic
MVCs agrees both with earlier research [12] on the same
muscle, and a study [24] of the vastus lateralis, albeit
only at 0.52 rad s⫺1. In contrast, most of the more recent
studies [14–16] reported that the EMG from knee exten-
sor muscles was greatest in the middle of the range of
motion for isokinetic MVCs, but offered no explanation
for this. It is unlikely that this contention is due to dif-
fering subject familiarisation, as both sets of studies
[12,15] have reported extensive periods of practice
before testing. However, it is possible that the way in
which subjects were instructed to perform the MVCs
254 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
Fig. 6. Mean (+SD) root mean square difference (RMSD) between
the isokinetic MVC method and the other normalisation methods for
concentric and eccentric phases of isotonic contractions at four differ-
ent forces.
could influence the EMG-angle relationship. In this
study subjects were instructed to pull against the lever
arm of the dynamometer as hard and fast as possible
throughout the entire range of motion. Details such as
these have not been presented in other studies [12,14–
16,24]. If a subject was simply instructed to perform an
MVC there might be a tendency to develop tension
gradually at the start and to relax at the end of the range
of motion.
The independence between EMG and angular velocity
of isokinetic MVCs again agrees with early findings [13]
for the same muscle. However, it does not agree with
more recent research [15,17–19] which has investigated
the same relationship for the knee extensor muscle group
and has discovered that EMG increased at higher angular
velocities of concentric isokinetic MVC. The fluctuation
of the EMGs from all isokinetic MVCs about the iso-
metric MVC value, shown in Fig. 4, indicates that the
biceps brachii was uniformly and similarly, if not maxi-
mally, activated during these “maximal” effort contrac-
tions. The slight decline in eccentric muscle activation
as the angular velocity increased, observed in the present
study, is unlikely to be the result of fatigue due to the
random selection of velocities and the absence of this
trend in the concentric data. These findings again contra-
dict those from studies of the knee extensor muscles,
which show greater muscle activation during the concen-
tric phase of isokinetic MVCs [15,17–19]. The reduction
in neural drive, during both eccentric contractions and
decreasing velocity concentric contractions, has pre-
viously been attributed to neural inhibitory mechanisms
[17–19], preventing utilisation of the full force-generat-
ing capacity of the muscle and protecting the joint and
surrounding tissues from injury [18]. The mechanisms
for this could involve feedback from joint receptors, free
nerve endings in the muscle, cutaneous receptors, pain
receptors and Golgi tendon organs [19]. The expected
difference in neural drive between the concentric and
eccentric phases of isokinetic MVCs was not observed
in highly skilled athletes who were used to performing
both types of contraction [17]. However, in the same
study, a difference was evident in a group of sedentary
subjects who were not used to performing such contrac-
tions. Subjects in the current study were not all highly
skilled athletes, but were extremely familiar with per-
forming the test contractions, having taken part in simi-
lar experiments in previous weeks. It is possible that the
subjects in the other knee extensor studies [15,18,19]
only received a limited amount of practice of eccentric
and high velocity concentric isokinetic MVCs before the
actual tests, which may be a reason for the reported
reduction in neural drive. The period of habituation
reported in those studies varies, but it is certainly a factor
worthy of further investigation.
4.2. Effect of normalisation method on the magnitude
of EMGs
As previously stated, and in agreement with previous
research [3], the magnitude of the outputs of the
Dynamic Mean and Dynamic Peak methods differed
 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
Table 1
Inter-subject coefficients of variation for each method of normalising EMGs for different forces and concentric (CON) and eccentric (ECC) mus-
cle actionsa
Normalisation method 50 N
CON ECC CON
None 0.62 0.59 0.54
Dynamic Mean 0.17 0.27 0.11
Dynamic Peak 0.25 0.34 0.28
Isometric MVC 0.34 0.53 0.25
Isokinetic MVC 0.32 0.53 0.24
a
The CVs presented in Table 1 were calculated using a single trial from each subject rather than an ensemble average as in other studies
referred to in the text [3,5,11].
both from each other and from the Isometric and Isoki-
netic MVC methods by virtue of the different denomi-
nators used in the normalisation equation. The magni-
tude of the outputs from the Isometric and Isokinetic
MVC methods were similar, as reinforced by the small
RMSDs shown in Fig. 6. This is unsurprising given the
uniform EMG-angle and angular velocity relationships
reported for the biceps in this and previous [12,13] stud-
ies, which were discussed above. The power of the stat-
istical tests and the validity of our findings would obvi-
ously be improved by increasing the number of subjects
and also assessing the reliability of the results. However,
the five subjects who did take part in this investigation
all exhibited similar, uniform EMG-angle and angular
velocity relationships. As a consequence of this all sub-
jects displayed a minimal difference between the output
of the Isometric MVC and Isokinetic MVC methods.
Had this study used knee extensor muscles, instead of
an elbow flexor muscle, a larger difference might have
been seen between the two normalisation methods,
because of the different EMG-angle and EMG-angular
velocity relationships that have previously been reported
for that muscle group [14–19].
Only two studies [4,20] have previously compared
EMGs normalised using the Isometric MVC Method
with methods similar to the Isokinetic MVC Method.
Unfortunately, neither study used the RMSD to compare
methods. Therefore, it is difficult to compare them
directly with the findings of the present study. However,
if the percentage difference method, as used by Mirka
[4], is used to compare the outputs from the Isokinetic
and Isometric MVC methods, values of between ⫺15%
and 18% are obtained for concentric and eccentric con-
tractions at all forces. These values are less than the
average 15 to 50% and maximum 68 to 689% differ-
ences reported by Mirka [4] for trunk muscles. The dis-
crepancy between the investigations could be due firstly
to the different muscles used and, secondly, because
Mirka [4] normalised EMGs from dynamic contractions
with EMGs from isometric MVCs taken at the same
trunk angle but not the same angular velocity. In partial
255
Force
100 N 150 N 200 N
ECC CON ECC CON ECC
0.65 0.37 0.56 0.38 0.49
0.17 0.10 0.14 0.11 0.11
0.41 0.15 0.30 0.07 0.19
0.36 0.09 0.27 0.18 0.21
0.30 0.12 0.20 0.26 0.10
agreement with Mirka [4], two subjects consistently pro-
duced higher outputs when the EMGs were normalised
using the Isometric MVC Method than when the Isoki-
netic MVC Method was used (represented by a negative
percentage difference). In comparison, Kellis and
Baltzopoulos [20] discovered that the output from a nor-
malisation method, similar to the Isokinetic MVC
Method, was significantly greater than the output from
the Isometric MVC Method, by up to 127%. Differences
between the two methods would be expected owing to
the non-uniform EMG-angle and angular velocity
relationships reported for the knee extensor muscles.
However, to produce differences of this magnitude, the
EMGs obtained from the isometric MVC must have
been, in some cases, twice as large as those obtained
from the isokinetic MVCs. This suggests that the isoki-
netic MVCs undertaken by Kellis and Baltzopoulos [20]
yielded EMGs that were far from maximal. Comparison
between studies must be viewed with caution as the
magnitude of the normalised EMG can be affected by
the method used to process the denominator of the nor-
malisation equation, as well as the nature of the denomi-
nator, especially when normalising EMGs from different
angular velocities of motion [11]. However, the findings
of this study are worthy of comparison with those of
Mirka [4] and Kellis and Baltzopoulos [20] as their
EMGs were also processed using the mean absolute
value, albeit with a smaller averaging window of 20 ms
and 10 ms respectively.
The assumption that the Isokinetic MVC Method pro-
vides the most accurate representation of the percentage
of the muscle’s maximal activation capacity required to
perform a particular task, may be challenged by evidence
which has frequently been used to criticise the Isometric
MVC Method. That is, isokinetic MVCs may also pro-
duce EMGs that are unreliable and do not represent the
maximal activation capacity of the muscle. The findings
from the present investigation cannot answer the first
criticism, as only one contraction was performed at each
angular velocity to prevent fatigue. In partial response to
the second criticism, Fig. 4 shows that, for some angular
256 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
velocities, the mean EMGs for both the concentric and
eccentric contractions were less than the EMG recorded
from the isometric MVC. Further evidence that the isoki-
netic and isometric MVCs were not truly maximal,
despite “maximal” effort being used, is given in Fig. 5,
which shows that the concentric EMGs normalised by
both methods were in excess of 100% for some subjects
when moving the 200 N load.
4.3. Effect of normalisation method on inter-subject
variability of EMGs
The use of normalisation methods similar to the
Dynamic Mean and Dynamic Peak methods has success-
fully reduced inter-subject variability [5,10]. The coef-
ficients of variation presented in Table 1 lend support to
the already strong evidence that use of the Dynamic
Mean Method reduces the inter-subject variability in
relation to other normalisation methods [3,5,11] and the
un-normalised EMG [5,11]. In agreement with Allison
et al. [11], the coefficient of variation was also generally
higher for eccentric contractions when the EMGs were
normalised using this method. Thus, if researchers or
clinicians wish to retain the homogeneity of task-specific
EMGs for a group of individuals, they should avoid use
of the Dynamic Peak and, in particular, the Dynamic
Mean normalisation methods, as suggested elsewhere
[3,11]. The Isometric MVC Method also reduced inter-
subject variability in relation to the un-normalised EMG,
as did the Isokinetic MVC Method. Therefore, even
these two methods falsely improved group homogeneity
albeit to a lesser degree. This is in comparison to Allison
et al. [11], who found that the Isometric MVC method
increased the coefficient of variation in relation to the
un-normalised EMGs. The disagreement is unlikely to
be due to the different methods of processing used by
Allison et al. [11] (root mean square) and ourselves
(mean absolute value), as these authors also reported that
the coefficient of variation was largely unaffected by the
way in which the isometric MVC was processed.
4.4. Ability of normalisation method to detect changes
in external force
The Isometric and Isokinetic MVC methods did reflect
the increase in EMG that occurred in response to the
three increments in external force. This supports the
findings of Allison et al. [11] that the Isometric MVC
method was able to distinguish between load and no-
load conditions, in both concentric and eccentric con-
tractions of the biceps brachii, regardless of the way in
which the EMG from the isometric MVC was processed.
Unlike the MVC methods, the Dynamic Mean and
Dynamic Peak methods are not designed to provide the
percentage of the maximal activation capacity of the
muscle required to perform the isotonic contractions.
Hence, it is unsurprising that the output of the Dynamic
Peak and, in particular, the Dynamic Mean normalisation
methods were unable to reflect the increase in EMG that
occurred in response to the increase in force. This dis-
agrees with the findings of Allison et al. [11] that nor-
malisation methods which used either the mean or the
peak EMG from the ensemble average were able to dis-
tinguish between load and no-load conditions for the
same muscle. In view of our findings, this is particularly
surprising considering that the difference between the
load and no-load conditions was only 2.3 kg. As pre-
viously stated, Allison et al. [11] reported that the magni-
tude of the normalised EMG may be affected by the way
in which the denominator of the normalisation equation
is processed. Thus, it is possible that this discrepancy
exists due to the different processing methods used by
Allison et al. [11] and ourselves. However, this is
unlikely as Allison et al. [11] found that processing
method did not affect the ability of the Isometric MVC
method to distinguish between load and no-load con-
ditions. In addition, by virtue of the nature of the
denominator used in their normalisation equations,
neither the Dynamic Peak nor Dynamic Mean method
would be expected to reflect changes in muscle acti-
vation levels between tasks.
The trend of the RMSD between the two MVC
methods to increase slightly as the external force
increased, in both concentric and eccentric contractions,
can be explained using the following simple hypothetical
example. With a 50 N external load, the magnitude of
the EMG at a particular angle and angular velocity dur-
ing an isotonic contraction is 0.5 mV. The magnitudes
of the EMG from an isokinetic MVC at the same angle
and angular velocity, and an isometric MVC at an arbi-
trary angle, are 3.2 mV and 3.0 mV respectively. If the
EMG from the 50 N trial is normalised using the Isoki-
netic MVC and Isometric MVC methods they produce
outputs of 15.6% and 16.6% respectively, a difference
of 1%. With a 200 N external load, the magnitude of
the EMG at the same angle and angular velocity rises
to 2.8 mV. If this EMG is normalised, an output of
87.5% is produced by the Isokinetic MVC Method and
93.3% by the Isometric MVC Method, a larger differ-
ence of 5.8%. Thus, regardless of the effect of the EMG-
angle and EMG-angular velocity relationships, the absol-
ute difference between the two methods will always be
greater for higher muscle activation.
5. Conclusions
The following conclusions were drawn from this
study:
1. Biceps brachii EMGs recorded during concentric
and eccentric isokinetic MVCs of the forearm
 A. Burden, R. Bartlett / Medical Engineering & Physics 21 (1999) 247–257
flexors were found to be largely unaffected by the
angle or angular velocity of flexion, or the type of
muscle action involved. This agreed with early find-
ings for the same muscle [12,13], but was not in
agreement with more recent work involving the knee
extensor muscle group [14–19].
2. The Isokinetic MVC and Isometric MVC methods
produced normalised EMGs which had a similar
magnitude of output, as a consequence of the uni-
form EMG-angle and EMG-angular velocity
relationships that were discovered for the biceps bra-
chii. This disagreed with recent investigations of the
trunk [4] and knee extensor [20] muscles that com-
pared normalisation methods which were similar to
the MVC methods.
3. Both of the MVC methods also displayed slightly
lower inter-subject coefficients of variation than
those observed in the un-normalised EMGs. By con-
trast, and in agreement with previous findings
[3,5,11], both the Dynamic Mean and the Dynamic
Peak methods displayed smaller coefficients of vari-
ation and, therefore, falsely improved group hom-
ogeneity.
4. Both the MVC methods were able to distinguish
between changes in muscle activation in response to
differing magnitudes of force. In comparison, both
Dynamic methods were unable to distinguish
between the same changes in force. Thus, the Iso-
metric or Isokinetic MVC, and not the Dynamic
Peak or Dynamic Mean, methods should be used to
normalise the amplitude of EMGs if the objectives
are to compare them between muscles, tasks and
individuals, or to retain the natural variation between
those individuals.
However, owing to the apparent differences in maximal
EMG-angle and angular velocity relationships between
muscles, recommendation of either MVC related nor-
malisation method is, at present, only advocated for the
biceps brachii. Future studies should, therefore, investi-
gate the EMG-angle and angular velocity relationship in
other muscles, and continue to evaluate and compare the
Isometric and Isokinetic MVC normalisation methods.
References
[1] De Luca CJ. The use of surface electromyography in biomechan-
ics. J Appl Biomechanics 1997;13:135–63.
[2] Mathiassen SE, Winkel J, Hagg GM. Normalization of surface
EMG amplitude from the upper trapezius muscle in ergonomic
studies — a review. J Electromyogr Kinesiol 1995;5:197–226.
[3] Knutson LM, Sodberg GL, Ballantyne BT, Clarke WR. A study
of various normalization procedures for within day electromyo-
graphic data. J Electromyogr Kinesiol 1994;4:47–59.
257
[4] Mirka GA. The quantification of EMG normalization error. Ergo-
nomics 1991;34:343–52.
[5] Yang JF, Winter DA. Electromyographic amplitude normaliz-
ation methods: improving their sensitivity as diagnostic tools in
gait analysis. Arch Phys Med Rehabil 1984;65:517–21.
[6] Arsenault AB, Winter DA, Marteniuk RG. Is there a “normal”
profile of EMG activity in gait? Med Biol Engng Comput
1986;24:337–43.
[7] Dubo HIC, Peat M, Winter DA, Quanbury AO, Hobson DA, Ste-
inke T, Reimer G. Electromyographic temporal analysis of nor-
mal gait: normal human locomotion. Arch Phys Med Rehabil
1976;57:415–20.
[8] Yang JF, Winter DA. Electromyography reliability in maximal
and submaximal isometric contractions. Arch Phys Med Rehabil
1983;64:417–20.
[9] Enoka RM, Fuglevand AJ. Neuromuscular basis of the maximum
voluntary force capacity of muscle. In: Grabiner M, editor. Cur-
rent issues in biomechanics. Champaign: Human Kinetics,
1993:215–35.
[10] Winter DA, Yack HJ. EMG profiles during normal human walk-
ing: stride-to-stride and inter-subject variability. Electromyogr
Clin Neurophysiol 1987;67:402–11.
[11] Allison GT, Marshall RN, Singer KP. EMG signal amplitude nor-
malization technique in stretch-shortening cycle movements. J
Electromyogr Kinesiol 1993;3:236–44.
[12] Komi PV, Buskirk ER. Effect of eccentric and concentric muscle
conditioning on tension and electrical activity of human muscle.
Ergonomics 1972;15:417–34.
[13] Komi PV. Relationship between muscle tension, EMG and velo-
city of contraction under concentric and eccentric work. In:
Desmedt JE, editor. Basel: Karger, 1973:597–605.
[14] Bobbert MF, Harlaar J. Evaluation of moment-angle curves in
isokinetic knee extension. Med Sci Sports Exerc 1992;25:251–9.
[15] Eloranta V, Komi PV. Function of the quadriceps femoris muscle
under maximal concentric and eccentric contractions. Electromy-
ogr Clin Neurophysiol 1980;20:159–74.
[16] Kellis E, Baltzopoulos V. Agonist and antagonist moment and
EMG-angle relationship during isokinetic eccentric and concen-
tric exercise. Isokin Exercise Sci 1996;6:79–87.
[17] Amiridis IG, Martin A, Morlon B, Martin L, Cometti G, Pousson
M, van Hoecke J. Co-activation and tension-regulating phenom-
ena during isokinetic knee extension in sedentary and highly
skilled humans. Eur J Appl Physiol 1996;73:149–56.
[18] Seger JY, Thortensen A. Muscle strength and myoelectric activity
in prepubertal and adult males and females. Eur J Appl Physiol
1994;69:81–7.
[19] Westing SH, Cresswell AG, Thorstensson A. Muscle activation
during maximal voluntary eccentric and concentric knee exten-
sion. Eur J Appl Physiol 1991;62:104–8.
[20] Kellis E, Baltzopoulos V. The effects of normalization method
on antagonistic activity patterns during eccentric and concentric
isokinetic knee extension and flexion. J Electromyogr Kinesiol
1996;6:235–45.
[21] Clarys JP, Cabri J. Electromyography and the study of sports
movements: a review. J Sports Sci 1993;11:379–448.
[22] Okamoto T, Tsutsumi H, Goto Y, Andrew P. A simple procedure
to attenuate artefacts in surface electrode recordings by painlessly
lowering skin impedance. Electromyogr Clin Neurophysiol
1987;27:173–6.
[23] Games PA, Howell JF. Pairwise multiple comparison procedures
with unequal n’s and/or variances: A Monte Carlo study. J Educat
Stat 1976;1:113–25.
[24] Ghori GMU, Donne B, Luckwill RG. Relationship between
torque and EMG activity of a knee extensor muscle during isoki-
netic concentric and eccentric actions. J Electromyogr Kinesiol
1995;5:109–15.