<span class="gmail_quote"><br></span>Christina -<br>
<br>
Under ideal conditions, extended ICA is able to separate line noise
into a small number of components. Tzyy-Ping recently reminded me that
this number might not be one, since buildings get three-phase power,
meaning some circuits may have line noise with different phases. Then
again, the electric field may change during the experiment (e.g. when
the elevator goes by, when AC units go on/off, etc.), and these may
have different phases, meaning that the (phasor) sums in each electrode
circuit may differ, and not be resolvable into one component. We have't
investigated this in the data, though - an interesting task would be to
find the phase at the 60-Hz (or 50-Hz) peak in each independent
component whose spectrum has such a peak. Are the phase values all
different or not? (Remember that there is also endogenous 60 Hz, so
averaging across time will be necessary to get a stable value).<br>
<br>
In general, we do not worry much about 60 Hz - unless it is so large
that it obscures the rest of the data! In that case, using
filtering to make it smaller may be a good idea. Actually, we are
planning to electrically isolate our new lab room (with
metal-containing wallpaper and copper mesh under the carpet), but this
is not an option for all, and may also require a very good building
ground to be effective.<br>
<br>
Scott Makeig<div><span class="e" id="q_110144a3d3a026c5_1"><br>
<br><br><div><span class="gmail_quote">On 1/10/07, <b class="gmail_sendername">Christina Karns</b> <<a href="mailto:ckarns@berkeley.edu" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">ckarns@berkeley.edu
</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>I'd like to hear some opinions about the best way to deal with line noise.<br>I have a relatively clean dataset where the power spectrum of the<br>continuous data shows relatively little 60Hz line noise (we use the
<br>Biosemi system which is less susceptible to line noise than some other<br>systems). Should I use a notch filter or ICA to separate out any line<br>noise contributions to the data?<br><br>1. One approach would be to use ICA to separate out the line noise.
<br>According the the EEGLAB manual, "ICA can separate out certain types of<br>artifacts -- only those associated with fixed scalp-amp projections. These<br>include eye movements and eye blinks, temporal muscle activity and line
<br>noise." Note that, "when a source distribution is sub-Gaussian (e.g., as<br>with line noise), the extended power spectrum and three event-related<br>time/frequency option of infomax ICA must be used to separate it."
<br><br>2. However a previous discussion on this list: (see<br><a href="http://sccn.ucsd.edu/pipermail/eeglablist/2004/000716.html" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">http://sccn.ucsd.edu/pipermail/eeglablist/2004/000716.html
</a>) suggests that<br>it would be better to notch filter the data and then perform ICA so "any
<br>artifact of independent origin is likely to occupy one degree of freedom<br>in ICA space to model that artifact. If you have too many, you are likely<br>to be left with no degrees of freedom for sources of interest."
<br><br>It seems that the advantage of using ICA to separate out the line noise is<br>that there may be neural processes that overlap with the 60Hz frequency<br>band that would be filtered out with the notch filter. In constrast, ICA
<br>should be able to distinguish these if they are independent.<br><br>Do you have any advice for me?<br><br>Thanks,<br><br>**********************************<br>Christina M. Karns<br>University of California, Berkeley<br>
***********************************<br>_______________________________________________<br>eeglablist mailing list <a href="mailto:eeglablist@sccn.ucsd.edu" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">
eeglablist@sccn.ucsd.edu</a><br>Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">
http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>To unsubscribe, send an empty email to <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">eeglablist-unsubscribe@sccn.ucsd.edu
</a><br><br></blockquote></div><br><br clear="all">
<br></span></div><span class="sg">-- <br>Scott
Makeig, Director and Research Scientist, Swartz Center for
Computational Neuroscience, Institute for Neural Computation,
University of California San Diego, La Jolla CA 92093-0961,
<a href="http://sccn.ucsd.edu/%7Escott" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">http://sccn.ucsd.edu/~scott</a>
</span><br clear="all"><br>-- <br>Scott Makeig, Director and Research
Scientist, Swartz Center for Computational Neuroscience, Institute for
Neural Computation, University of California San Diego, La Jolla CA
92093-0961, <a href="http://sccn.ucsd.edu/~scott">http://sccn.ucsd.edu/~scott</a>