>From PAUL GUY University of Waterloo.
"Here's a few things you could check: is there any DC (more than
a few millivolts) across the differential inputs to your pre-amps?
(measure with a battery powered handheld digital voltmetetr - DON'T
use a power line connected instrument). If there is, then change to
another kind of electrode. The DC potential can be 'modulated' by
extreme movements. Sometimes an added bit of electrode paste can help.
We've had different branfs of electrodes that gave anywhere from less
than a millivolt to some with 200 or more millivolts DC. Needless to
say, the ones with the high DC were a real pain. Currently we're
using "Medi-Trace" pellet electrodes (part ECE-1801) from Graphic
Controls Canada, they have quite low DC offsets.
In your skin preparation, are you using fairly pure alcohols to clean
the skin? If you are, then they will dry up the natural skin moisture,
and drastically raise the impedance. High skin impedance will cause a
lot of artifact conditions. We use a 50% water and ethanol for skin
prep. Typical skin impedances are less than 50 kilohms, usually around
10 kilohms. Using pure alcohols can raise it to 200 kilohms or more.
You can measure it ...but be VERY careful. DON'T use a line operated
meter it's an extremely hazardous thing to do. Some meters can use
quite high voltages and currents to measure resistance. A lot of the
new portable voltmeters will work OK, but check with your electronics/
safety people first.
Make sure your wiring is not being yanked, or bent in a small radius.
I assume that the wire is shielded properly, and has low noise and
tribo- electric properties. Also ensure that there is not a lot of
movement or pulling at the place where the EMG leads plug into your
Is there a reference or 'ground' electrode (not power ground)? Is your
system fully isolated or are you using telemetry to isolate the subject
from the power system?
Is your subject wearing any clothing that will generate quite high
voltages? THese aren't problems with a telemetered system, but can
play havoc with so called 'isolated systems'. There's enough stray
capacitance, resistance in some isolated measurement systems that static
and 60 Hz fields can still be a problem.
Your 10Hz low frequency limit in your system could be a problem. What
order filter is it? i.e. how fast does it drop off at the low end. I'd
say a second order is the minimum required, we've found that in some
cases (for ruinning) it was better to use 15-20 Hz, with a 2nd or 3rd
order. If your data has been converted, and resides on a computer, you
can use a fourier type filter to get rid of everything below certain
frequencies. They are quite effective at reducing heart signals from
EMG measurements in the trunk area, and work superbly at eliminating
60 Hz pick-up.
E-mail - paul@gaitlab1.uwaterloo.ca
"I suspect there is a difference in impedance between the skin and the
electrodes between those trials that do or do not show an artefact.
Often the artefact can be reduced by extremely thorough skin prepara-
tion. Put a multi-meter (BATTERY POWERED) across your electrodes while
they are in place. You must get less than 1000 Ohms resistance. This is
only a crude measure of electrode impedance but it gives you a guide.
You'll probably notice that those trials giving most artefact will
coincide with the highest resistance.
E-mail - p.sinclair@cchs.su.edu.au
"I would suggest using active (on-site pre-amplified) electrodes which
I experienced as fairly stable against motion artefacts. There is still
a need for good skin preparation of the recording sites. We shave to
remove hair and the upper skin layer, clean with 70% alcohol (a burn-
ing sensation is a good sign for having shaved thoroughly) and apply
electrode gel by rubbing it into the skin. Then we fix the bipolar
active electrode with double adhesive washers and secure it by an
additional tape. Tapping on the electrode and moving the wires is used
to test for artefact. If these movements show on the raw signal we
repeat the above process but usually we are successful the first time.
E-mail - diro@sirius.medizin.umi-ulm.de
"Because we wanted to analyze the signal during movements (fast wheel-
chair propulsion) we had to solve the problem of movement artefacts
caused by changes of the position of the electrodes relative to the
muscle (skin movements) and by changes in the inter-electrode distance.
Extensive research on analytical movements of the knee, elbow and trunk,
have brought us to an algorithm we are using now for almost ten years.
It is a recursive formula which I will try to indicate in this message.
A parameter called activity and indicated by "A(i)", is calculated as a
function of time and derived from a series of EMG samples "E(i)". The
algorithm calculates a value of A which relates very linearly to the
amount of static force (Pearson correlation coefficients of typically
0.98). It is clear that this value is then used to relate to muscle
force (or moment in the joints) in dynamic situations, taking the con-
traction speed, mode (eccentric, concentric), pre-load etc.. into
account. To simplify the type-writing of the formula I will use here
a PASCAL like expression:
A(i+1) : = 0.975*A(i)+sqrt (abs (E(i+1)-E(i))).
Constant holds, as an example, for sample frequency of 500 Hz. As you
will have noticed, activity A(i+1) depends on the amount of activity
that is left from a previous time instant (A(i)) and on the new acti-
vity in the next interval derived from the difference between two
subsequent EMG values. The rationale between this formula is explained
in a paper which is under preparation at the moment and should be sub-
mitted in April. The formula itself is presented and illustrated in
trwo papers of which I could send a copy by fax. The formula is almost
insensitive to movement artefacts since it uses the difference between
subsequent EMG samples. Those differences are only marginally influenced
by relatively slow changes in the base-line of the EMG signal.
Finally the recursive expression assumes an initial value for A. In
most situations this is not known. We have proven however that values
calculated for time instances later than approximately 0.5 sec. from
the initial value are almost insensitive for the initial value of A.
For this reason the problem of the initial value can always be solved
by sampling the EMG signal at least one second earlier than the time
where "A" is of interest.
E-mail - Arthur.Spaepen@flok.kuleuven.ac.be
Try another type of electrode e.g., Medicotest (DK), E-10-VS (a ECG
electrode for prematures, "VS" stands for Very Small). The electrode
is NOT pre-gelled which means you can fill the active part with gelly
and get a better connection to the skin. Make sure that the 10cm
electrode wire is applicated in a winding with enough space to move
without pulling the electrode. Secure the electrode pair with a
"net-shirt" of Surgifix (a fixing material used in e.g. surgical
stomi operations, produced by FRA Production Italy.)
Increase the lower filter limit to 20 Hz (doesn't affect the EMG
spectrum too much. Most of the movement artefacts are low frequency
1-5 Hz.
E-mail - larsbrun@nioh.se
"I think you should try to use pre-amplified electrodes (so-called)
active electrodes) with, for instance, a gain of 10. With a pre-
amplification near the recorded area your signal/noise ratio is much
better, avoiding the negative ijnfluence of mechanical artefacts. We
have used these procedures for ballistic movements without any prob-
lems with mechanical artifacts coming from the movement of cables.
E-mail - np24ab@mail.telepac.pt
"My experience with this is that the wires move in the magnetic field
of the earth and generate small voltages. THe size of the voltages will
depend on the orientation of the wire with the field which might ex-
plain the variable results. One way to reduce the relative intensity of
this artifact is to reduce the impedance measured at the electrode -
skin interface. It should be less than 10 Kohms and hopefully less than
1 Kohm.
E-mail - Richard.Smith@cchs.su.edu.au

TO BE CONTINUED....................