PDA

View Full Version : Linear Envelope method summary.



Jaime B. Zamora S.
07-09-2000, 08:35 AM
Thanks to all answers.

----------------------------------------------------------------------------
-----
Original message:

Hello all coleagues.

I saw in many papers a method for sEMG signal processing called linear
envelope, but it's doesn't appear explicitly in the papers. If anyone knows
how this method works please let me know.

Thanks in advance.
----------------------------------------------------------------------------
--


Answers:
--------------------------------------------------------------------------
Hi Jaime
You first full wave rectify the raw EMG signal, then you low pass
filter the signal and the frequency cut off used will depend on what
you are going to do with the data. For example, gait work looking
at patterns of muscle activity will use a 6 Hz cut off, whereas EMG
to force processing may use a one Hz and studies that wish to
retain the timing information will use a higher cut off i.e. up to
100Hz. The linear envelop and moving average are often the same
processing. This processing can be done analogue or digital.
Winter is probably one of the best references. There are several
papers in the early 80s, plus his book and more recently a chapter
in Kumar and Mital Electromyography in Ergonomics, Taylor and
Frances.
Cheryl K.
----------------------------------------------------------------------------
try
Winter, D.
Biomechanics and Motor Control of Human Movement
Wiley-Interscience Publication, John Wiley and Sons, Inc
New York, Chisceter, Brisbane, Toronto, Singapore
1990

p 204 in 2nd ed. linear envelope is explained in detail

good luck,
lise


Lise Worthen MS, BFA
Biomechanist and Lab Coordinator
Soar Research @ LIU
122 Ashland Place #1A
Brooklyn, NY 11201
718-246-6378
718-246-6383 fax
lworthen@liu.edu
------------------------------------------------------------------
See chapter 8 on EMG in Winter's 1990 text, but basically it is the result
of passing the rectified EMG signal through a low pass filter with a low
cut-off frequency. Winter has suggested a compromise of 3 Hz. This provides
a moving average of the signal that can be better related to kinetic data
because it follows a similar trend to the tension at the tendon in both
patterning and timing. Be careful of the use of the term in the literature.

regards,
Brad McFadyen
----------------------------------------------------------------------------
---------
hi Jamie,

There are many descriptions of Linear envelope in the EMG literature , but
briefly you take your signal and full wave rectify it then low pass filter
it. This results in an "envelope" representation of the data in the time
domain- the concept of an "envelope" is that the representation encapsulates
or envelopes the signal - the degree to which this does this accurately of
course, depends on the original sample rate and the low pass filter type and
cut-off . similar in net result but different in process is to take a RMS
(root mean Square) approach to the data.

hope this helps,

also try this website for some basic info:
http://www.delsys.com/emg_articles/emg_articles.html
Kevin
----------------------------------------------------------------------------
-------
Jaime:

Creating a linear envelope involves two general steps:

1. Full-wave rectify the data. This is done simply by taking the absolute
value of the raw EMG signal
2. Low-pass filter the full-wave rectified data. If temporal alignment with
other data are important (heel strike, toe off, kinematic data, etc.), be
sure to use a zero-lag filter (like a second order Butterworth filter run in
the forward and backward directions). The cutoff frequencies used for
low-pass filtering are usually described in the literature.

Depending on how you will use these data, I would also recommend correcting
for dc-bias before step 1. This can be done by either a) [better] applying a
high-pass filter to the raw EMG data; or b) [okay] subtracting the average
raw EMG value from the instantaneous raw EMG data
(dc_bias_corrected_raw_EMG[i] = raw[i] - mean_raw_EMG).

Hope that was helpful.

Peter
___________________________________________
Peter F. Vint, Ph.D.
Research Integrations, Inc.
9180 S. Kyrene Rd., Suite 117
Tempe, AZ 85284
Phone: (480) 893-1600 x14
Fax: (480) 893-0602
e-mail: peter.vint@researchintegrations.com
----------------------------------------------------------------------------
------
A "linear envelope" of a band-passed EMG signal (1 Hz - 1 kHz) can be
created by;

1. Full wave rectification (FWR) of the signal (assuming the signal has no
offset)
2. Low pass (LP) filtering (cutoff between 4-50 Hz)

Both the FWR and the LP filter can be performed digitally. The FWR being an
absolute valuye of the signal. The low pass filter can be performed using a
recursive, Butterworth digital filter. The recursive feature passes the data
thru forward and then reversed to "remove" the phase lag produced by the
"forward" pass of the filter.



Dean Kriellaars, Ph.D.
Associate Professor
Human Performance Lab
School of Medical Rehabilitation, University of Manitoba
Voice: (204) 787-2289 Fax: (204) 787-1227
----------------------------------------------------------------------------
-----------
Linear envelope is rectification followed by low pass filter (3-10 Hz
cutoff) of the EMG signal.

Warren Darling
----------------------------------------------------------------------------
----
Jaime:

The text "Biomechanics of Human Movement" by David Winter is a classic
source for the technique of creating a linear envelope. Basically, with EMG
the steps are 1) rectify the raw EMG signal; 2) apply a low pass filter at
an appropriate cut-off to the rectified data.
Regards, Steve


Steven T. McCaw, Ph.D
Professor, Biomechanics
Dept of HPER
5120 Illinois State University
Normal, IL 61790-5120

Phone: 309-438-3804
Fax: 309-438-5559

----------------------------------------------------------------------------
----------

Its simple: take the absolute value of the signal (called full-wave
rectification) then apply a low-pass filter with a cutoff between 4 an 6
hertz. Viola, you have the linear envelope EMG.

D. Gordon E. Robertson, Ph.D.
Assistant Director (Graduate Studies) and
Professor of Biomechanics
School of Human Kinetics, University of Ottawa
Ottawa, ON, Canada, K1N 6N5
mailto:dger@uottawa.ca
+1-613-562-5800 Off: 4227 Lab: 4246 Fax: +1-613-562-5149
Webpage: http://www.health.uottawa.ca/biomech/lab
----------------------------------------------------------------------------
Hi Jamie

Basically the procedure to generate a linear envelope is as follows:

1). Filter the signal using a bandpass filter with a lower cut-off
frequency of 20 Hz and an upper cut-off frequency of 500 Hz
(assuming the EMG was collected at 1000 Hz or greater).

2). Full wave rectify the filtered signal (negative values become
positive).

3). Low pass filter the rectified signal with a cut-off frequency of 3
Hz.

The result is the linear envelope of the EMG signal.

The cut-off frequencies I have used are a guide only. You may
wish to change these to suit your own purposes.

Good luck

Pete
Peter Mills BExSc (Hons)
PhD Candidate
School of Physiotherapy and Exercise Science
Griffith University
PMB50 Gold Coast Mail Centre
Queensland 9726
Ph: +61 7 5594 8390
Fax: +61 7 5594 8674
Mobile: 0410 503 415
Email: p.mills@mailbox.gu.edu.au

----------------------------------------------------------------------------
--

Linear envelope processing consists of full-wave rectification (absolute
value of each sample so that the entire signal is positive) followed by
low-pass filtering (choice of cut-off frequency, yours).

Samuel Lee, M.Sc.
Research Engineer
Hospital for Special Surgery
Department of Biomechanics & Biomaterials
New York, N.Y. 10021 U.S.A.
(212) 774-2382
LeeS@HSS.edu

----------------------------------------------------------------------------
------------

Jamie,

Linear envelope detection involves the following steps.

1. rectify the EMG by taking the abolute value.
2. detrend the EMG by removing any slope in the baseline or DC offset.
3. low-pass filter the EMG with a cutoff frequency no greater than 10 Hz

The result is supposed to be a EMG/time curve that follows the Force/time
curve.

-d.g.
David A. Gabriel, Ph.D.
Assistant Professor
Biomechanics
Physical Education Department
Brock University
St. Catharines, Ontario, CANADA
L2S 3A1

Phone: 905-688-5550 ext.4362
Fax: 905-688-5550 ext. 4104

e-mail: dgabriel@arnie.pec.brocku.ca

----------------------------------------------------------------------------
------

Most of these articles assume you know what it is, I realize.
I went through a whole slew of them, and a linear envelope seemed to be
a fancy way of saying, draw a line that outlines the general activity.
This can be done a number of ways
1) a running average: average the data points across 60 or so ms, if not
more. Obviously, the higher the averaging window, the smoother the
data. It should be done on rectified data; ie, positive half of the
sine wave, otherwise, the negetive troughs will create an average at
zero.
2) write a routine that reads all the peaks, and plots it.
3) a filtering program, like 4th order butterworth, should have a
similar smoothing effect as the first method. You would probably need to
have a cutoff frequency as low as 10 Hz, depending on the data you want
to see, the sampled frequency, and the pre-filtering cutoff frequencies
you used.
I noticed that this simply shrinks the amplitude of your data more than
smoothing, until a certain point.
Good luck -
Janina Wilen

--
:::::::::::::::::::::::::::::::::::::::::::::::::: ::::::::::
Janina Wilen
BioMechanical Engineer, B.S.
Kessler Medical Research and Rehabilitation Corp
Human Performance and Movement Analysis Laboratory
West Orange, NJ 07052
973-243-6903
fax: 973-243-6984
jwilen@kmrrec.org
http://www.kmrrec.org

----------------------------------------------------------------------------
-----


Regards,

http://www.ing.puc.cl/icmtesis/magister/jzamora/index_esp.html
Jaime B. Zamora S.
Laboratorio DAMAC.
Departamento de Ingeniería Mecánica y Metalúrgica.
Pontificia Universidad Católica de Chile.


BEGIN:VCARD
VERSION:2.1
N:Zamora S.;Jaime;B.
FN:Jaime B. Zamora S.
ORG:PUC;DAMAC Lab.
TITLE:M. Sc. student
TEL;WORK;VOICE:56-2-6864409
TEL;CELL;VOICE:56-09-4296556
TEL;WORK;FAX:56-2-6865828
ADR;WORK:;DAMAC Lab.;Vicuña Mackenna 4860;Santiago;Santiago;;CHILE
LABEL;WORK;ENCODING=QUOTED-PRINTABLE:DAMAC Lab.=0D=0AVicu=F1a Mackenna 4860=0D=0ASantiago, Santiago=0D=0ACHILE
ADR;HOME:;;;;;;CHILE
LABEL;HOME:CHILE
X-WAB-GENDER:2
URL:http://www.ing.puc.cl/icmtesis/magister/jzamora/jzamora.htm
EMAIL;PREF;INTERNET:jzamoras@ing.puc.cl
EMAIL;INTERNET:jzamoras@puc.cl
REV:20000709T223506Z
END:VCARD