View Full Version : Summary of Responses (Normalizing Methods of the EMG)

02-12-2010, 04:58 AM
Dear Biomech-L subscribers,

Thank you to everyone who responded to my request for advice on
"Normalizing methods of the EMG".

The original posting is below, followed by the full replies I received.

Warm regards,
Asimakis K. Kanellopoulos,
MSc (Distinction) in Bioengineering

Lecturer in Physiotherapy
Technological Educational Institute of Lamia, Greece

Postgraduate Researcher in Biomechanics Analysis Room
Bioengineering Unit, Engineering Faculty
University of Strathclyde
Glasgow, Scotland, UK

cell phone (GR) +30 6973 513040

--------Original Message-------------------------
Dear Colleagues,

I am interested in normalizing my surface EMG results (Quadriceps,
both static and
dynamic trials). I have been told
a) to divide the RMS by the MVIC (peak MVIC or mean?- not clear),
b) to divide the entire signal by the MVIC (peak MVIC or mean?-not
clear) and then take the RMS
c) to divide the entire signal by the entire MVIC signal and then
take the RMS, etc.

Unfortunately, I cannot retrieve in detail any specific method about this issue
from the literature. Does anyone know the best method to do this, or
advice me about a paper dealing with this in detail (explaining for
example pros-cons etc)?

Thank You very much!

PS: MVIC=Maximal Voluntary Isometric Contraction

Dear Asimakis,

You may take a look at my method to normalize it to a functional task,
which is more comparable. In most cases, the MVC of MVIC signal is too
large, and your collected data stay within 0-5% of the MVIC signal
only. So the resolution is not fully utilized.

Fong, D.T.P., (2008) Lower extremity preventive measures for
slips-Joint moments and myoelectric analysis.Ergonomics,Vol.51,No.12,

Daniel T.P. FONG, PhD, FISBS,
Research Assistant Professor, Department of Orthopaedics and Traumatology,
Prince of Wales Hospital, Faculty of Medicine, The Chinese University
of Hong Kong, Hong Kong, China.
Phone: (852) 26323535 / 96020151; Fax: (852) 26463020; Email:
Department of Orthopaedics and Traumatology: http://www.ort.cuhk.edu.hk
Director, Sport Performance and Biomechanics Laboratory:
Deputy Director, MSc/PgD program in Sports Medicine and Health
Science: http://www.cuhk.edu.hk/whoctr/MSc
Board of Directors, International Society on Biomechanics in Sports:
Commission Member, Hong Kong Association of Sports Medicine and Sports
Science: http://www.hkasmss.org.hk
Managing Editor, SMARTT journal: http://www.smarttjournal.com
Managing Editor, JOSR journal: http://www.josr-online.com


Dear Mr. Kanellopoulos,

I think the MVIC is one of the method for normalization the EMG data.
Maybe you can consider another method: normalized the EMG activity by
the mean value. (see attached)

Barela, A.M.F., et al.(2006) Biomechanical characteristics of adults
walking in shallow water and on land.Journal of Electromyography and


Chi-Wen LUNG MSc, PhD
Research Fellow
Institute of Intelligent Machines, Chinese Academy of Sciences
Science Peninsula, Hefei, Anhui 230031, China,
Tel +86 551-5591104
Mobile +86 139-6667-4473
dragon1234@gmail.com http://www.iim.ac.cn

Dr. Kanellopoulos,

There are different ways to normalize an EMG signal. The simplest way
is to take the highest value of the MVIC and divide your signal with
this value.

I will suggest you to check the following link from delsys:


you will find everything that you need in this web page.

Take care,

Panagiotis Koutakis
Path Integration Project Director
Peripheral Arterial Disease Research Assistant
University of Nebraska at Omaha
Nebraska Biomechanics Core Facility
Fellow of Alexander S. Onassis Public Benefit Foundation
(402) 208-8933


Dear Asimaki,

Typically, we use the following procedure:

We identify the highest activation area of the EMG in both the
submaximal contraction and the maximal contraction (~0.5 s; depending
of course of the length of your signal). Then to identify the
normalized activation we do the following: (RMS of submaximal / RMS of
maximal) x 100.

So it is closest to the first option you are presenting. As far as
references, I would look for any Enoka RM papers that use EMG
normalization (closest to your task; most likely papers with BL
Tracy). Possibly also review papers by Merletti, Farina, or Enoka.
Feel free to check some of my papers on my website as I have used
similar procedures in some of my work.

I am not familiar with the options b and d .

Hope this helps. Please let me know if you need anything else.

Yeia Xara!

Evangelos A. Christou Ph.D.
Assistant Professor
158U Read Building
Department of Health and Kinesiology
Texas A&M University
College Station TX 77843-4243

Phone: 979-862-3089
email: eachristou@hlkn.tamu.edu
website: http://neuromuscularphysiology.tamu.edu/



Unfortunately, I cannot retrieve in detail any specific method
about this issue
from the literature. Does anyone know the best method to do this, or
advice me about a paper dealing with this in detail (explaining for
example pros-cons etc)?

I would say that the problem is: there is no general agreement and the
discussion will probably never be solved because a BEST METHOD might
not even exist. Several approaches have been used and appear

Therefore, you might want to try if the different approaches yield
differing results or might even show the same trends (as could be
expected). Therefore, I would advise to select one method and clearly
describe it so that others can understand how you got to your
normalized EMG.


Dieter Rosenbaum...

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~
Prof. Dr. Dieter Rosenbaum
Funktionsbereich Bewegungsanalytik - Motion Analysis Lab
Klinik fuer Allgemeine Orthopaedie - Orthopaedic Department
Universitaetsklinikum Muenster - University Hospital Muenster
Domagkstr. 3
D-48149 Muenster Germany

Fon: +49 (0)251 - 8352970
Fax: +49 (0)251 - 8352993
Email: diro@uni-muenster.de
Web: http://motionlab.klinikum.uni-muenster.de www.kidfoot.de
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~


Dear Asimakis,

a good reference is the ABC of EMG by Peter Konrad. Just google it and you
will be able to download the pdf.

Good luck and best wishes,


Max Feltham PhD
Postdoctoral Researcher
Movement Science Group
School of Life Sciences
Oxford Brookes University
Headington Campus
Gipsy Lane
United Kingdom

T: +44 (0)1865 483272
F: +44 (0)1865 483242
E: mfeltham@brookes.ac.uk


Have you tried...


Their 'Use of Surface EMG in Biomechanics' recommends...

'When normalizing the amplitude of the EMG signal, do so at values less than
80% MVC. Above this
level, the EMG signal and the force (torque) are exceptionally unstable and
do not provide a suitable
reference point.'


'Measure the MVC by choosing the greatest value of three consecutive
attempts at reaching the maximal
value, with a rest period of at least 2 min between contractions. Let the
subject choose his/her own
force rate to reach maximal value. The posture of the subject should be
similar if not identical to that
assumed during the actual test. Repeat this measurement each time the
experimental conditions



Does this help?

J Electromyogr Kinesiol. 2009 Sep 14. [Epub ahead of print]

Methodological aspects of SEMG recordings for force estimation - A
tutorial and review.
Staudenmann D, Roeleveld K, Stegeman DF, van Dieën JH.

Department of Integrative Physiology, Neurophysiology of Movement
Laboratory, University of Colorado, Boulder, CO, USA.

Insight into the magnitude of muscle forces is important in
biomechanics research, for example because muscle forces are the main
determinants of joint loading. Unfortunately muscle forces cannot be
calculated directly and can only be measured using invasive
procedures. Therefore, estimates of muscle force based on surface EMG
measurements are frequently used. This review discusses the problems
associated with surface EMG in muscle force estimation and the
solutions that novel methodological developments provide to this
problem. First, some basic aspects of muscle activity and EMG are
reviewed and related to EMG amplitude estimation. The main
methodological issues in EMG amplitude estimation are precision and
representativeness. Lack of precision arises directly from the
stochastic nature of the EMG signal as the summation of a series of
randomly occurring polyphasic motor unit potentials and the resulting
random constructive and destructive (phase cancellation) superimposit
ions. Representativeness is an issue due the structural and functional
heterogeneity of muscles. Novel methods, i.e. multi-channel monopolar
EMG and high-pass filtering or whitening of conventional bipolar EMG
allow substantially less variable estimates of the EMG amplitude and
yield better estimates of muscle force by (1) reducing effects of
phase cancellation, and (2) adequate representation of the
heterogeneous activity of motor units within a muscle. With such
methods, highly accurate predictions of force, even of the minute
force fluctuations that occur during an isometric and isotonic
contraction have been achieved. For dynamic contractions, EMG-based
force estimates are confounded by the effects of muscle length and
contraction velocity on force producing capacity. These contractions
require EMG amplitude estimates to be combined with modeling of muscle
contraction dynamics to achieve valid force predictions.

PMID: 19758823 [PubMed - as supplied by publisher]



Matthijs Tuijt, PhD Student

Academic Centre for Dentistry Amsterdam (ACTA)
Dept. Functional Anatomy / Oral Kinesiology
Meibergdreef 9 Room A1.24
1105 AZ Amsterdam
The Netherlands

Telephone +31-(0)20-5665355
Fax +31-(0)20-5669524
Email m.tuijt@amc.uva.nl
Website www.acta.nl



My background is not really in EMG processing, but I use it in my
research. I've attached some documents that have helped me, but I too
have never found anything addressing which normalization method is
One more processing alternative not listed below (and the one I'm
actually using right now) is to normalize to the subject-specific
maximum value for the task being performed. This gives you a reasonable
normalization without having to go through the trouble of an MVC or
MVIC. As usual, the "best" method usually depends on what questions
you're asking, but this one works for me.

1. (a pdf technical note having the following info)
In investigations where the force/torque is correlated to the EMG it
is common to normalize the force/torque and
its respective EMG, relative to the values at maximal voluntary
contraction (MVC) Obtaining the best estimate
of MVC from subjects requires some preliminary training. Without
training, the MVC could be as much as 20-30
% less than that obtained after appropriate training and lead to
incorrect conclusions or interpretation of data.
Estimates of MVC may be performed in different conditions that should
be described (e.g. with/without
biofeedback, position of the subject, condition of the joint proximal
to the one of interest, etc.)
Standards for Reporting EMG Data
Ó 1999 by International Society of Electrophysiology and Kinesiology
Normalising the force/torque with respect to its MVC value is commonly
performed with MVC as 100 % of
force/torque, and other force levels are expressed as the appropriate
% of MVC. Similarly, the EMG associated
with 100 % MVC is designated as 100 % and fractions thereof. Both
force/torque and EMG normalization
should include other relevant information such as joint angle(s)
and/or muscle length(s) in isometric contractions,
and range of joint angle, muscle length, velocity of shortening or
lengthening and load applied for non-isometric
In sum, the following information should be provided when normalizing data:
· how subjects were trained to obtain MVC
· joint angle and/or muscle length
· conditions and angles of adjoining joint, e.g., for studies on
elbow flexion, the condition of the wrist and
shoulder joints should be provided
· rate of rise of force
· velocity of shortening or lengthening
· ranges of joint angle or muscle length in non-isometric contraction
· load applied in non-isometric contractions"

2. Konrad, P. "The ABC of EMG", e-book (can be googled)

3. Clancy,E.A.Sampling, noise-reduction and amplitude estimation
issues in surface electromyography.Journal of Electromyography and
Kinesiology 12 (2002) 1–16

Brian Schulz, Ph.D.
Biomechanics Researcher
HSR&D/RR&D Center of Excellence, Maximizing Rehabilitation Outcomes
James A. Haley Veterans' Medical Center
8900 Grand Oak Circle, Room 149
Tampa, FL 33637-1022
Phone: (813) 558-3944
Fax: (813) 558-7691


Dear Asimakis,

for my experience, normalization of EMG is still a controversial
subject and it strongly depends on the task you are performing. I had
the same problem with normalization in cycling, since in literature
often a MVIC is used and pedaling is a dynamic task. Therefore, I
proposed and used a new isokinetic pedaling test.

I think that the key, at least in dynamic exercises, is that the
reference contraction has to be the most similar as possible to the
contraction you want to normalize (joint position etc), and of course
you have to analyze them in the same way.

Please find attached a couple of recent papers about this topic, and
don't hesitate to contact me if I can be helpful.

1. Norcross, M.F. et al. Reliability and interpretation of single leg
stance and maximum voluntary isometric contraction methods of
electromyography normalization. Journal of Electromyography and
Kinesiology, in press

2. Ferna´ndez-Pen˜a, E. et al. A maximal isokinetic pedalling exercise
for EMG normalization in cycling. Journal of Electromyography and
Kinesiology 19 (2009) e162–e170.


Eneko Fernández, PhD


Hi Asimakis,

I saw this email of yours from BIOMCH-L. I hope I am not responding too late.
Regarding normalization, there is no fixed method. People use
different ways to normalize EMG and there is no best way to do it.
As in your methods you will anyway explain how you normalized the
data. Remember you are using normalization to compare across
(technical you are keeping the same process across your data) that is
why you need not to worry about which method is best. Just use what
you feel easy or good.

You can use mean or Peak value in MVICs. I personally use Peak value
(just a I feel it easy that way). Usually reviewers will not question
on normalization, if mention how you did it in your methods sections.

Good luck,

Research Engineer,
Rehabilitation Institute of Chicago.
Chicago, IL, USA

Hi Asimakis,

As you know there are many ways that you can normalise EMG data from
MVICs. To be honest, it probably doesn’t matter too much whether your
divide the RMS by the MVIC or the whole signal by the MVIC. I think
the key is to make sure that whichever approach you adopt is stable
and consistent. You might like to pilot some of the different
approaches that you’ve listed below and determine which is the most
stable - that is, chose the one that gives your data less variability
between trials and participants.

My research is focused on gait trials. We calculated the normalised
EMG by dividing a 600ms RMS window (MVIC or sub-max MVIC) by specific
phases on the gait cycle (both peak and RMS). In addition to this, I
recommend using a dynamic/sub-max MVIC approach to normalise gait
data, as it consistently produced lower variability.

I’ve attached our paper just published in J Biomech related to these issues.

Kind regards,


1. Murlay, GS. et al. Reliability of lower limb electromyography
during overground walking:A comparison of maximal-and sub-maximal
normalisation techniques. Journal of Biomechanics, in press.

George S. Murley

Lecturer and First Year Podiatry Coordinator
Department of Podiatry & Musculoskeletal Research Centre
Division of Allied Health
La Trobe University

ph +61 3 9479 5834 e-mail g.murley@latrobe.edu.au web
latrobe.edu.au/podiatry & latrobe.edu.au/mrc


Hello Asimakis,

Here is what I do:

For my MVC trial with the greatest amplitude and within 5% of another
MVC trial, I find the peak force, center a 500 ms window about this
time, and take the RMS amplitude as the "Max RMS".

I then normalized my data from other contractions to this value. I
commonly divide my contraction into equal duration epochs (thirds or
fifths) and then use a repeated-measures ANOVA to assess if the
amplitude of the signal changes with time.

Let me know what you think,

- Mike

Mike Pascoe, M.S.
Doctoral Candidate
Integrative Physiology
University of Colorado
Boulder, CO 80309-0354, USA

T (303) 492-4975
F (303) 492-6778