Biosensor Array Galvanic Skin Response Spurious Input

Advice on dealing with spurious input from Biosensor Array Galvanic Skin Response by Sean Montgomery

---------- Forwarded message ----------
Date: Fri, 29 Apr 2011 14:42:57 -0400
From: Sean Montgomery <xxx@gmail.com>
To: Drew Fustini <xxx@xxx.com>
Subject: Re: GSR: eliminating spurious changes?

Unless you're a hardware guru, software is usually easier to quickly try out
a bunch of analyses and see what might work. You could use the Arduinoscope
processing library (similar to one of the ones we published in Make
Magazine) to plot the data along with some simple analyses and detections.
-s



On Fri, Apr 29, 2011 at 2:32 PM, Drew Fustini <xxx@xxx.com> wrote:

Hi Sean,

Thanks for your fast and detailed response.  We've been learning a lot
during the short build period and existing, well-documented projects like
yours have been a big help.

I like your suggestion of measuring numerous locations to reject spurious
changes.  I'm imaging that would be done best in software.  Is that what you
were thinking as well?

Thanks!
Drew


---------- Forwarded message ----------
Date: Fri, 29 Apr 2011 14:09:40 -0400
From: Sean Montgomery <xxx@xxx.com>
To: Drew Fustini <xxx@xxx.com>
Subject: Re: GSR: eliminating spurious changes?

Hi Drew,

Ah, you've found the difficulties of bio-sensing... separating signal from
noise.

The problem you're experiencing is that when you move, you change the amount
of contact between your finger and the electrode, thereby changing the
resistance of the junction... exactly what the GSR is measuring. You can try
to find a part of the body that is less likely to move during your
behaviours of interest. I've gotten decent GSR from the wrist and not much
success from the forehead, but fingers are the best I've found. You
can measure GSR from numerous locations and reject spurious changes that
only occur on one or two of the locations. You can also try to characterize
the waveshape of real GSR responses and reject resistance changes that don't
fit that criteria.

Biosensing has huge potential, but is not without its hurdles. Some things
are easier to measure like heart rate where you can more easily reject noise
because of the stereotyped electrical signature. On the far end of the
spectrum is EEG, where you're separating tiny electrical potentials in a sea
of electrical noise and THEN trying to understand how those handful of
electrical traces relate the hundreds of billions of neurons in the brain
and the unknown (large) number of ongoing cognitive processes.

Generally the trick is to understand your signal, what's creating it, what
sorts of limitations that puts on your experimental paradigm. It's good to
think about what kinds of workarounds are possible (as above), but also
think about how you can limit the scope of your question to fit the
attributes of the experimental tools that are available.

Sorry no easy fix for ya. I'll be interested to know what you come up with.

Cheers,
Sean



On Fri, Apr 29, 2011 at 1:34 PM, Drew Fustini <xxxx@xxxx.com> wrote:

Hi Sean,

I'm having a problem with finger cuffs which you note on the Truth
Wristband page:

"Note that physically moving the electrodes can create spurious changes in
the resistance measured across the plates and contaminate our measurement.
There are ways to work around this, but it's not completely trivial."

I built the two-finger coper foil velcro cuffs and also the single-finger
cuff from the wristband kit.  The trouble I have from both is flexing or
moving one's fingers generates a signal that I can't differentiate from
actual GSR response (like poking my leg with scissors).

Our biosensor array is design to be worn on one's body so having the hands
motionless for accurate GSR isn't ideal.  I'd really appreciate if you'd
have any advise on how to eliminate spurious GSR readings due to movement.

BTW, your Make co-author, Ira, pleasantly surprised me with a call to my
desk phone to introduce an interesting biosensor data aggregation project
he's getting organized :)

Thanks!
Drew