Date: Fri, 26 Jul 1996 10:39:19 -0600
Reply-To: Roger James
Sender: Biomechanics and Movement Science listserver
From: Roger James
Subject: Summary of Replies: Normalizing EMG data
Dear Biomch-L colleagues,
Thank you to everyone who responded to my
posting of July 22, 1996 regarding EMG
normalization procedures. Rather than
appending all responses I'll attempt to
highlight the most relevant information. I have
organized this summary into the following
1. List of contributors
2. Summary of recommendations
3. Take-home lesson
4. List of suggested references
5. Original posting
LIST OF CONTRIBUTORS (14 total)
Garry Allison, Curtin Univ. of Technology
Kelly Cole, Univ. of Iowa
J. D. Cooke, Univ. of Western Ontario
David Davis, Loughborough Univ.
Marcos Duarte, Universidade de Sao Paulo-Brasil
Mark Grabiner, Cleveland Clinic Foundation
Steve Ingram, NASA
Don Kirkendall, Duke Univ. (?)
Anne Mannion, Univ. of Bristol
Mal McHugh, Lenox Hill Hospital
Rich Shiavi, Vanderbilt Univ. (?)
Drew Smith, Univ. of Toronto
David Wong, Univ. of Waterloo
Wiebren Zijlstra, Univ. Hospital Groningen
SUMMARY OF RECOMMENDATIONS
The following responses are the highlights of
relevant information from each contributor.
Note that some responses were shortened and/or
paraphrased in order to save space, but I tried
to preserve the primary content and intent.
From Garry Allison (email@example.com):
MVC (isometric) is the standard but there are
problems which may influence the quality of
the data & statistical inferences:
1. normalise if the process is not reliable (other
points in no set order)
2. MVC need repeated sessions (at least two) to
establish a reliable estimate.
3. Submaximal is more reliable than MVC - especially
in pathological groups.
4. Using Coefficient of Variation as an assessment tool
for judging the effectiveness of a normalisation
technique is at the possible expense of removing 'true
5. From #4. normalisaing may (does) alter the statistical
power of the analysis.
6. In studies where you are investigating neuromusclular
control/synergistic patterns etc. following an
intervention, it is very difficult, if not impossible, to
have a normalisation procedure which is independent
of the intervention program.
7. MVC (isometric) is often used to normalise for
concentric and eccentric actions (SSC).
8. Due to non-linear force amplitude factors...submaximal
normalisation should be at a similar amplitude as the
task you are assessing.
From Kelly Cole (firstname.lastname@example.org):
Dick Brand and colleagues at U. Iowa have studied
reliability of EMG signals during gait in ACL-deficient
patients....They employed exacting quantitative and
statistical methods that may be of relevance. His
papers should be traceable through MedLine or
similar search; he can be contacted at
From J. D. Cooke (email@example.com):
Some years ago when looking at movements about
the elbow we utilised a different "biological assay"
for normalizaiton (see Brown,S.H., Cooke, J.D.
(1981): Amplitude and instruction dependent
modulation of movement-related electromyogram
activity in humans. J Physiol (Lond) 316, 97-107).
This method was based on the observation that (at
least at lower force levels) the integrated surface
EMG varies linearly with isometric force.
Normalization within or across subjects required
simple linear scaling.
From Marcos Duarte (firstname.lastname@example.org):
In order to normalize EMG data there are
three main procedures:
1. normalize to a maximum voluntary isometric
contraction (MVIC) EMG
2. normalize to peak dynamic EMG
3. normalize to mean dynamic EMG
In the last two procedures you normalize by the
peak/mean value of each subject's EMG data.
From Mark Grabiner (email@example.com):
You may want to consider expressing the amplitude
of voluntary EMG to the amplitude of the maximum
M-wave. The maximum M-wave can be elicited by
delivering a single supramaximal electrical pulse to
the motor nerve.
From Mal McHugh (firstname.lastname@example.org):
If there is good agreement in normal subjects between
right and left, or dominant and nondominant legs then
you could normalize the ACL data to the uninvolved
leg. Even if ACL patients were able to tolerate an
MVC it may not be correct to normalize to this MVC
given the significant post-op quadriceps inhibition in
From Rich Shiavi (email@example.com):
Myself and Winter have found that normalizing the
linear envelope by its average value minimizes the
variability of the envelopes across experiments and
enables comparisons across subjects.
By the way, ISEK has just created a committee to
update the EMG standards.
From Drew Smith (firstname.lastname@example.org):
Dave Winter has normalised some of his EMG data
by taking the average voltage of the linear envelope
during the contraction and dividing the curve
by that value.
When I was a student working with Dave, another
of his students, Sylvia Ounpuu, normalised her EMG
data against known isometric joint torques (via
weights). For some muscles, this required some rather
creative thinking. She works in Connecticut, but I
don't have her email address. Try contacting
someone in the North American Society of Gait and
Clinical Movement Analysis.
Other individuals who's comments are not given
above either asked specifically that their message not be
included in a public summary or primarily contributed to the
The consensus seems to be that maximum voluntary
contractions (MVC) are not necessarily ideal for
normalizing EMG data under all circumstances. MVCs
are typically acquired during isometric muscle actions.
It would seem to make little sense to normalize a concentric
or eccentric DYNAMIC activity to the isometric MVC
value. Additionally, MVCs are often obtained by isolating
joints and/or muscles and by requiring motions dissimilar to
the activity of real interest (e.g., gait, jumping, landing,
cycling, etc.). Therefore, there appears to be at least three
major decisions that must be made in choosing an
appropriate normalization procedure: 1. the type or form
of the activity used as the criterion; 2. the intensity,
workload, or effort involved in the criterion activity versus
the functional or experimental activity; 3. the actual
numerical EMG parameter used as the criterion. The
responses given above and the references given below
address some or all of these points. Relative to my
original question: "are there criterion values other than
MVCs (e.g., a submax contraction at a standardized
workload or effort, etc.) that may be appropriate and that
will allow valid and meaningful comparisons among
subjects?", the answer appears to be YES. A submax
contraction may be appropriate and some suggest that
it may be even more reliable than a MVC. We'll
explore this possibility for our application.
LIST OF SUGGESTED REFERENCES (given as received)
1. Allison, Marshall & Singer (1993). EMG signal amplitude
normalisation technique in stretch-shortening cycle
movements. J Electromyography and Kinesiology, 4 (4),
2. Brown, S. H., & Cooke, J. D. (1981). Amplitude and
instruction dependent modulation of movement-related
electromyogram activities in humans. J Physiol (Lond),
3. Edgerton, V. R,. et al (1996). Theoretical basis for
patterning EMG amplitudes to assess muscle dysfunction.
Med Sci Sports and Exerc, 28 (6), 744-751.
4. Knutson, Soderberg, Ballantyne, & Clarke (1994). A
study of various normalization procedures for within
day electromyographic data. J Electromyography and
Kinesiology, 4 (1), 47-59.
5. Neuman, P., Norman, R., & Wells, R. (incomplete
6. Potvin, Norman, & Wells (1990). A field method for
continuous estimation of dynamic compressive forces on
the L4/L5 disc during the performance of repetitive
industrial tasks. 23rd Annual Congress of the Human
Factors Association of Canada.
7. Shiavi & Zhang (1992). J Orthop Res, 10, 226-236.
8. Wells, R., Norman, R., & Neumann, P. (incomplete
9. Winter, D. A., & Yack, H. J. (1987). EMG profiles
during normal walking: stride-to-stride and inter-subject
variability. Electroenc and Clinical Neurophysiology,
67 (5), 402-411.
10. Yang & Winter (1983). Archive of Phys Med & Rehab,
11. Yang, Winter, & Eng (1984). Electromyographic
amplitude normalization Methods: Improving their
sensitivity as diagnostic tools in gait analysis. Archive
of Phys Med & Rehab, 65, 517-521.
12. Zhang & Shiavi (1991). IEEE Trans Biomed
Engineering, 38, 777-784.
>A graduate student of mine is working on a thesis and has the
>need to collect and analyze EMG information for "normal"
>individuals and post-op "ACL" patients. In order to compare
>muscle activity among subjects and between groups, the EMG
>data must be normalized in order to account for "...individual
>differences in subcutaneous fat, muscle geometry, and other
>variances..." (Soderberg & Cook, 1984). Traditionally, the
>criterion used to normalize EMG has been a maximum
>voluntary contraction, MVC, of the involved muscles, as
>suggested by a variety of sources including Basmajian &
>DeLuca (1985) and Soderberg & Cook (1984). While the
>use of MVCs to normalize EMG data have been recommended,
>the MVC value is often difficult to obtain and/or difficult
>or inappropriate to relate to the functional activity of interest.
>In the case of the current study, we are concerned that post-op
>ACL patients may not be able to tolerate the forces involved
>in a MVC. The question is: are there criterion values other than
>MVCs (e.g., a submax contraction at a standardized workload
>or effort, etc.) that may be appropriate and that will allow
>valid and meaningful comparisons among subjects? A search of
>the Biomch-L archives and a perusal of approximately 500
>titles available on Medline and WorldCat have been little help
>relative to this specific question. Additionally, the International
>Society of Electrophysiological Kinesiology (ISEK) EMG
>Standards (reprinted in Dainty & Norman, 1987) did not
>really address the normalization problem. I welcome the
>opinions, experiences, and recommendations of this group and
>will be happy to post a summary of replies.
>Basmajian, J. V. & DeLuca, C. J. (1985). Muscles Alive.
>Baltimore: Williams & Wilkins.
>Dainty, D. A. & Norman, R. W. (Eds.) (1987). Standardizing
>Biomechanical Testing in Sport. Champaign, Illinois:
>Soderberg, G. L. & Cook, T. M. (1984). Electromyography in
>Biomechanics. Physical Therapy, 64(12), 1813-1820.
End of Summary.
Thanks again to everyone who contributed and additional
comments are still welcome.
:: C. Roger James, Ph.D.
:: Assistant Professor, Biomechanics
:: Department of Health, Physical Education & Recreation
:: Center for Sports Health & Human Performance
:: Box 43011 Texas Tech University
:: Lubbock, Texas USA 79404-3011
:: (806) 742-3371 Phone
:: (806) 742-1688 FAX
:: Email: email@example.com
:: [Find us on the Web--http://www.ttu.edu/~hper]
Return to PE 506 page
BACK TO GORDON'S HOME PAGE