Re: software specs

From: peterson_at_d...
Date: 2001-05-07 02:11:01


Rob--I was very impressed to find out how much this software can
already do. I had no suspicion that you were this far along, having
watched people like the Lexicor programmers bumbling along with no
major design changes in years, even though none of their stuff will
run under Windows so you have to screw around trying to free up all
kinds of memory under the 640k barrier, etc. and then can barely get
it to run on most machines. It doesn't like the IBM or Compaq
parallel ports although nobody knows why, etc. They don't even have
any printer drivers more modern than an ancient one for a 12 year old
H-P Paintjet.

--- In buildcheapeeg_at_yahoogroups.com, "Rob Sacks" <editor_at_realization.org> wrote:
> Hi Jim,
>
> > If you could make it put out displays like the original Mind
Mirror,
> > that might be enough for a beginning, for a first release.
>
> ElectricGuru already does a MindMirror-type display.
> (However the magnitudes are weighted differently, so
> high frequencies appear to be squashed inward compared
> to what they look like on Cade's machine.)
>
> There's now a screenshot showing the MindMirror display
> at http://www.realization.org/page/doc0/doc0086.htm.
> The illustration shows .5 Hz bands but they can be set
> narrower or wider. (As they get narrower, the epoch
> gets longer.) The page takes a while to download because
> the illustration is large and high-resolution.
>
> I agree about the games and sounds.
>
> > The software might permit the user to define EEG frequency bands
> > other than the classic Delta, Theta, Alpha and Beta...
>
> The current version of ElectricGuru does this.
>
> > It would also be neat if you could specify various Boolean
> > statements as conditions for feedback. E.g. tone1 if (SMR>x AND
> > Theta <y) or maybe symmetry training with a condition like tone1
if
> > ABS(rightalpha-leftalpha)<x where tone1 is the specified
feedback
> > tone.
>
> This is an appealing suggestion because it's probably
> the easiest way for me to implement user-programmable
> conditions. And it's a lot more elegant than menus. But
> I think most users would find it frightening and frustrating
> and difficult.

So, again--canned stuff and pretty "slider" interfaces for the
dilettantes, Boole for the rest of us. You could maybe make a few $$
doing training workshops for the logically challenged.
>
> I agree about canned protocols.
>
> > Whether using a canned session or an individually designed one,
it
> > should be easy to change the numeric values of the thresholds so
you
> > can easily adjust the difficulty levels during the session.
>
> How about just a single slider for raising or lowering the
> difficulty? It would work like this: if you set it at 60,
> the software would adjust all the threshholds so you are getting
> reward signals on 60 percent of the updates. You could also
> choose whether the software should readjust the threshholds
> continuously, thus keeping your rewards happening 60 percent
> of the time, or whether the adjustment should take place only
> when you manually click on "readjust threshholds." This second
> option would ensure that the person can sense the improvement
> taking place.

I think the slider should either come with a switchable display, one
for difficulty expressed as a percent, and one for difficulty as an
amplitude level. If you're just going to pick one or the other, go
for amplitude. I have serious doubts whether learning would be as
efficient under most conditions with a continuously variable
difficulty level.

> >
> > One would also want to be able to name and save training
protocols
> > for future use.
>
> This is already in the current version of ElectricGuru.
>
> > The feedback for any condition might be given whenever the
specified
> > condition is met or exceeded as averaged across a one-second
epoch.
> > Two seconds might be workable.
>
> The current version allows you to set the averaging to any
> amount of time you want down to the minimum required by
> the FFT. I have found that anything shorter than second
> is unusable because the reward signals are too erratic and
> "jumpy". This has made me extremely skeptical about
> the claims being made by some EEG manufacturers about
> their super-fast digital filters. If you have to average
> the output of those filters over a fairly long time anyway,
> what difference does it make how fast they are? I notice
> that Lexicor only uses FFT in its software.
>
"Some EEG manufacturers" beng Margaret Ayres, no doubt. I agree
about the technical impossibility of providing super-fast feedback
like she claims, and I've heard just about everybody in the field, at
least the equipment guys like Rex Hartzell, Frank Deitz, Adam Crane
and John Gilbert say the same thing. Furthermore, there is no
particular reason to think that there is any advantage to having
feedback that fast. The old conditioning literature showed a latency
of 1 or 2 seconds to be just about ideal, so I think this whole fast
feedback thing is a red herring.

I think that Lexicor uses FFT where it has to but also has some built
in notch filters.

> > The system should be accurate enough
> > so that no more than 5% or so of erroneous feedback is given.
>
> I don't know how to measure this.

Neither do I other than maybe by comparing it to some piece of highly
accurate equipment somewhere. I pulled this idea out of a hat.

> > Also, the software should provide the capability of tracking
> > amplitudes of the selected bands across the session.
>
> I agree. Excel output (actually, ASCII output which can be
> read by Excel) would be easy and flexible.
>
> > Would it be difficult to set the machine up to train coherence?
This
> > is when your right and left hemispheres not only match in the
> > dominant frequency, but are also in phase.
>
> ElectricGuru already does that kind of coherence -- you can
> set a threshhold on the number of degrees of phase difference.
> (It also does a Pearson correlation.)
>
> However, I have the impression that when people talk
> about coherence and synchrony, they don't necessarily
> mean this type of algorithm.
>
There does seem to be quite a bit of confusion in this area, even to
the point of there not being a standard or consistent terminology.

> One reason why you might want to test in a different
> way is to allow for one site to "drive" another.
> In such a case, the phase difference might be
> large, but it would remain relatively unchanging.
>
I take your point.

> At the time I stopped working on ElectricGuru, I was trying
> to find out whether there is any "standard" way of testing
> for coherence or synchrony or whatever the official term
> may be.
>
> > Now for a truly crazy request. Can it be made to produce
coherence
> > in two brains at the same time? With electrodes at the
> > corrresponding sites on the two heads, say at Cz or at Pz, where
> > alpha is often strongest.
>
> ElectricGuru can do that now, provided that the EEG
> machine can do it. I would think the BrainMaster could
> do it, although perhaps the two people would need to
> hold hands to maintain an identical reference voltage.
> However, I haven't tried it.

Fun to play with, though.

>
> Rob



This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:29 BST