Tweet
Thanks for the replies I received regarding my query. I had many replies
from list serv members saything they had nothing to contribute but were very
interested in our summary.
Again, thanks to all!
Sujani
Original query:
I am Sujani,a research engineer at the ADAM Center at Long Island
University, NY. The ADAM Center is a small laboratory that does research on
human movement. To date, we have focused on 3D kinematics and force
platform data capture and analysis.
Currently, I have started looking at the surface EMG data that we have
collected. We are particularly interested in onset and offset EMG
timing and would like to develop a computerized algorithm that allows us to
evaluate differing filters, threshold criteria (SD from baseline, etc),
samples assessed, etc. I particularly enjoyed the articles by Hodges
and Bui (1996): ³A comparison of computer-based methods for the
determination of onset of muscle contraction using electromyography² in
Electroencephalography and Neurophysiology.
I was wondering if I could get the latest updates and details on EMG
onset determination. As I was browsing the web, I came across a lot of
literature. It looks like each person uses a different value for
threhold, frequency for filtering,sample rate etc. What is the best way to
determine the onset?
I found out that some work has been done in MATLAB in this regard.
Before I start to write a new program in Labview, I was wondering if anyone
know about any shareware that may be available? Any ideas or
suggestions are welcome.
*******************************
Responses:
Paolo Bonato developed a method that he published in IEEE (A statistical
method for the measurement of muscle activation intervals from surface
myoelectric signal during gait. IEEE Trans Biomed Eng. 1998
Mar;45(3):287-99.). It develops an optimal criterion, based on the
statistics of the EMG signal and background noise. I had coded it in MATLAB
(enclosed --- the comments are limited, but I used the terms from his
article). PLEASE retest if you use the m-file --- I did the work a year or
two ago. In the end, we decided to manually annotate ensemble profiles,
since our work used cyclic activity (normal gait). Annotating individual
muscle actions can be rather challenging.
You might also look at some work by Dario Farina (A fast and reliable
technique for muscle activity detection from surface EMG signals. IEEE Trans
Biomed Eng. 2003 Mar;50(3):316-23.) This work is another approach.
Best of luck,
Ted Clancy
ted@ece.wpi.edu
******************************
Mega EMG system's MegaWin software has the onset analysis integrated in our
software. For further information, please contact our USA distributor, Peak
Performance Inc, Centennial, CO as follows:
Peak Performance Technologies, Inc.
7388 South Revere Parkway, Suite 901
Centennial, CO 80112
Mr Larry Scheirman
Mr Doug Reinke
Mr George Miller
Mr Christian Reist
Tel. +1-303-799 8686
Tel. 1-800-PIK-PEAK (USA only)
Fax. +1-303-799 8690
lscheirman@peakperform.com (Mr Scheirman)
dreinke@peakperform.com (Mr Reinke)
gmiller@peakperform.com (Mr Miller)
creist@2peakperform.com (Mr Reist)
Best regards
Mega Electronics Ltd
Kari Tiihonen
Export Manager
kari.tiihonen@meltd.fi
******************************
HI,
I'm a grad student in Kinesiology in Canada. I think, so far, we don't have
a gold standard for EMG onset threshold. To my knowledge, it all depends on
how neat is your data. In my project, I collect the EMG from gastro and
soleus and I used Baseline mean+3S.D to determine the onset. I saw couple of
literature used 4S.D. If you find any conlusion regrading it, could you let
me know? Since that's part of my thesis I need to write. Thanks!
Tony
church1026@yahoo.ca
******************************
In the following recently published paper a matlab application is described
Roetenberg D, Buurke JH, Veltink PH, Forner Cordero A, Hermens HJ.
Surface electromyography analysis for variable gait.
Gait Posture. 2003 Oct;18(2):109-17.
Priv. Doz. Dr. Dieter Rosenbaum
Funktionsbereich Bewegungsanalytik
Klinik und Poliklinik fuer Allgemeine Orthopaedie
Universitaetsklinikum Muenster
Domagkstr. 3
D-48149 Muenster
Fon: +49 (0)251 - 8352970
Fax: +49 (0)251 - 8352993
Email: diro@uni-muenster.de
http://medweb.uni-muenster.de/institute/orth/motionlab
******************************
Hi Sujani,
You are facing a "conflicting task". I discussed some of that in a
recent paper:
Morey-Klapsing G, Arampatzis A, Bruggemann GP. Choosing EMG parameters:
comparison of different onset determination algorithms and EMG integrals
in a joint stability study. Clinical Biomechanics. 2004, 19(2): 196-201
In the paper you will get several useful references.
Some advices from my side ... just test several filters and you will see...
1. Most filters introduce a time lag.
2. You can avoid this by filtering forwards and backwards, but you will
see that this widens the curve on the time axis.
3. I found (as suggested to me by At Hof) that the median filter pretty
much preserves the rises and falls, I use a 7 points back and forth
median filter (sampling at 1000 Hz).
4. It might be useful to include a further criterion in addition to only
signal amplitude. This would be above the threshold for at least some
time.
5. I would say (even if I didn't that way in my paper) that it is
preferable to take the mean and SD during a "zero" measurement ...
subject sitting or better lying flat and relaxed ... than whilst
standing, as backround actrivity might vary a lot.
In general I still prefer the integral of the signal in a fixed, self
defined interval comprising the interesting zone (for example during the
last 50 or 100 ms prior to ground contact). This will provide more
useful information and is not subject to the criticism of onset time
determination.
However for some issues Onset times are what we really want. And when
differences are big enough (lets say above 15 ms) and background
activity is determined in a relaxed condition this should be possible.
One more paper that might be useful:
Bonato, P.; D'Alessio, T; Knaflitz, M. A statistical method for the
measurement of muscle activation intervals from surface myoelectric
signal during gait. IEEE Transactions on Biomedical Engineering. 1988,
45(3), 287-299.
Good luck,
Gaspar
morey@dshs-koeln.de
******************************
Sujani,
For onset detection you can use the AGLR algorithm described in the
following article.
Objective motor response onset detection in surface myoelectric signals.
Staude G, Wolf W. Medical Engineering & Physics 21, pp 449-467, 1999.
Used for gait:
Surface electromyography analysis for variable gait
D. Roetenberg, J.H. Buurke, P.H. Veltink, A. Forner-Cordero, H.J. Hermens
Gait & Posture, Vol 18/2, pp 109-117, 2003
In those articles you will find information on threhold, frequency for
filtering,sample rate etc.
Daniel
D.Roetenberg@utwente.nl
******************************
-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
Please consider posting your message to the Biomch-L Web-based
Discussion Forum: http://movement-analysis.com/biomch_l
-----------------------------------------------------------------
from list serv members saything they had nothing to contribute but were very
interested in our summary.
Again, thanks to all!
Sujani
Original query:
I am Sujani,a research engineer at the ADAM Center at Long Island
University, NY. The ADAM Center is a small laboratory that does research on
human movement. To date, we have focused on 3D kinematics and force
platform data capture and analysis.
Currently, I have started looking at the surface EMG data that we have
collected. We are particularly interested in onset and offset EMG
timing and would like to develop a computerized algorithm that allows us to
evaluate differing filters, threshold criteria (SD from baseline, etc),
samples assessed, etc. I particularly enjoyed the articles by Hodges
and Bui (1996): ³A comparison of computer-based methods for the
determination of onset of muscle contraction using electromyography² in
Electroencephalography and Neurophysiology.
I was wondering if I could get the latest updates and details on EMG
onset determination. As I was browsing the web, I came across a lot of
literature. It looks like each person uses a different value for
threhold, frequency for filtering,sample rate etc. What is the best way to
determine the onset?
I found out that some work has been done in MATLAB in this regard.
Before I start to write a new program in Labview, I was wondering if anyone
know about any shareware that may be available? Any ideas or
suggestions are welcome.
*******************************
Responses:
Paolo Bonato developed a method that he published in IEEE (A statistical
method for the measurement of muscle activation intervals from surface
myoelectric signal during gait. IEEE Trans Biomed Eng. 1998
Mar;45(3):287-99.). It develops an optimal criterion, based on the
statistics of the EMG signal and background noise. I had coded it in MATLAB
(enclosed --- the comments are limited, but I used the terms from his
article). PLEASE retest if you use the m-file --- I did the work a year or
two ago. In the end, we decided to manually annotate ensemble profiles,
since our work used cyclic activity (normal gait). Annotating individual
muscle actions can be rather challenging.
You might also look at some work by Dario Farina (A fast and reliable
technique for muscle activity detection from surface EMG signals. IEEE Trans
Biomed Eng. 2003 Mar;50(3):316-23.) This work is another approach.
Best of luck,
Ted Clancy
ted@ece.wpi.edu
******************************
Mega EMG system's MegaWin software has the onset analysis integrated in our
software. For further information, please contact our USA distributor, Peak
Performance Inc, Centennial, CO as follows:
Peak Performance Technologies, Inc.
7388 South Revere Parkway, Suite 901
Centennial, CO 80112
Mr Larry Scheirman
Mr Doug Reinke
Mr George Miller
Mr Christian Reist
Tel. +1-303-799 8686
Tel. 1-800-PIK-PEAK (USA only)
Fax. +1-303-799 8690
lscheirman@peakperform.com (Mr Scheirman)
dreinke@peakperform.com (Mr Reinke)
gmiller@peakperform.com (Mr Miller)
creist@2peakperform.com (Mr Reist)
Best regards
Mega Electronics Ltd
Kari Tiihonen
Export Manager
kari.tiihonen@meltd.fi
******************************
HI,
I'm a grad student in Kinesiology in Canada. I think, so far, we don't have
a gold standard for EMG onset threshold. To my knowledge, it all depends on
how neat is your data. In my project, I collect the EMG from gastro and
soleus and I used Baseline mean+3S.D to determine the onset. I saw couple of
literature used 4S.D. If you find any conlusion regrading it, could you let
me know? Since that's part of my thesis I need to write. Thanks!
Tony
church1026@yahoo.ca
******************************
In the following recently published paper a matlab application is described
Roetenberg D, Buurke JH, Veltink PH, Forner Cordero A, Hermens HJ.
Surface electromyography analysis for variable gait.
Gait Posture. 2003 Oct;18(2):109-17.
Priv. Doz. Dr. Dieter Rosenbaum
Funktionsbereich Bewegungsanalytik
Klinik und Poliklinik fuer Allgemeine Orthopaedie
Universitaetsklinikum Muenster
Domagkstr. 3
D-48149 Muenster
Fon: +49 (0)251 - 8352970
Fax: +49 (0)251 - 8352993
Email: diro@uni-muenster.de
http://medweb.uni-muenster.de/institute/orth/motionlab
******************************
Hi Sujani,
You are facing a "conflicting task". I discussed some of that in a
recent paper:
Morey-Klapsing G, Arampatzis A, Bruggemann GP. Choosing EMG parameters:
comparison of different onset determination algorithms and EMG integrals
in a joint stability study. Clinical Biomechanics. 2004, 19(2): 196-201
In the paper you will get several useful references.
Some advices from my side ... just test several filters and you will see...
1. Most filters introduce a time lag.
2. You can avoid this by filtering forwards and backwards, but you will
see that this widens the curve on the time axis.
3. I found (as suggested to me by At Hof) that the median filter pretty
much preserves the rises and falls, I use a 7 points back and forth
median filter (sampling at 1000 Hz).
4. It might be useful to include a further criterion in addition to only
signal amplitude. This would be above the threshold for at least some
time.
5. I would say (even if I didn't that way in my paper) that it is
preferable to take the mean and SD during a "zero" measurement ...
subject sitting or better lying flat and relaxed ... than whilst
standing, as backround actrivity might vary a lot.
In general I still prefer the integral of the signal in a fixed, self
defined interval comprising the interesting zone (for example during the
last 50 or 100 ms prior to ground contact). This will provide more
useful information and is not subject to the criticism of onset time
determination.
However for some issues Onset times are what we really want. And when
differences are big enough (lets say above 15 ms) and background
activity is determined in a relaxed condition this should be possible.
One more paper that might be useful:
Bonato, P.; D'Alessio, T; Knaflitz, M. A statistical method for the
measurement of muscle activation intervals from surface myoelectric
signal during gait. IEEE Transactions on Biomedical Engineering. 1988,
45(3), 287-299.
Good luck,
Gaspar
morey@dshs-koeln.de
******************************
Sujani,
For onset detection you can use the AGLR algorithm described in the
following article.
Objective motor response onset detection in surface myoelectric signals.
Staude G, Wolf W. Medical Engineering & Physics 21, pp 449-467, 1999.
Used for gait:
Surface electromyography analysis for variable gait
D. Roetenberg, J.H. Buurke, P.H. Veltink, A. Forner-Cordero, H.J. Hermens
Gait & Posture, Vol 18/2, pp 109-117, 2003
In those articles you will find information on threhold, frequency for
filtering,sample rate etc.
Daniel
D.Roetenberg@utwente.nl
******************************
-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
Please consider posting your message to the Biomch-L Web-based
Discussion Forum: http://movement-analysis.com/biomch_l
-----------------------------------------------------------------