From: jpc@tauon.ph.unimelb.edu.au (John Costella)
Subject: Re: TECH: Neural Interfacing
Date: Sun, 13 Dec 92 14:35:59 EST


> From: chadwell@utkvx3.utk.edu (Chadwell, Leonard)
> 
> In article <1992Dec12.041658.16932@u.washington.edu>, dstampe@psych.toronto.e
>  (Dave Stampe) writes...
> 
> >There seems to be a lot on misinformation about neural connectivity
> >going around, so let me add mine (B-{))
> >
> >First, let's survey the techniques used so far: the up/down computers with
> >pattern recognition and scalp electrodes, multi-electrode EEG, direct
> >cortical contacts, nerve interface chips, and high-resolution NMR.
> >
> [stuff deleted]
> 
> Sorry, Dave, but L. Pinneo at SRI DID use EEG systems to do cortical wave
> pattern matching to qive qualified thought pattern detection.  Running on a P
> in 1974, and a skull cap with electrodes (no shaving or such required), the
> cursor on screen could be issued 7 commmands by thought:  UP,DOWN,LEFT,RIGHT,
> SLOW,FAST,STOP.  The main limitations on the system were the RAM (around 32K)
> and the speed of the processor.  The commands were merely thought, and the
> system could accurately recognize the commands on 60% of the people who were
> tested on the system, and with modification to pattern matching code, could
> also match those people.  With the progress made in raw computing power and
> memory capacity, more commands could be recognized with greater accuracy.

If this could be brought up to speed today, for a reasonable price,
then it could be useful for just what is mentioned here: navigating.
No more point, press, or whatever to move around, just think of the
direction you want to move. You'd have to add FORWARD and BACKWARDS
to the commands for VR, but I'm sure that's not too bad. All other
aspects of the VR hardware would still be the same.

You'd have to deal with all the mental activity generated by just
being in the virtual world to start with; maybe that would wipe out
the ``movement'' info from noise. But at least it might be feasible,
especially with the meaty processing power around today.

But, IMHO, from the info in the posts supplied it's pretty clear that 
this type of app, namely, acting as a physical transducer *from* the
participant, with a low rate of info (bits per second) is about all
that non-intrusive `neural interfaces' will do in the next 10 years. 
I don't think that you'll be able to plug a 80687 maths co-processor 
into your brain real soon. :) I'm not even sure that the brain could 
handle too many more input devices ... aren't all its slots full? :) 
(People with a loss of one or more senses excluded, of course.) Your
best bet is to use the existing inputs (e.g. eyes) and just interface
with them as efficiently as possible (like as in `impedance matched').
Things like the retinal scanner seem to be about the optimal way
to do that (any `ear-drum scanners' out there yet? :), if it turns
out to be cost-effective and light.

John

----------------------------------------------------------------------------
John P. Costella              School of Physics, The University of Melbourne
jpc@tauon.ph.unimelb.edu.au         Tel: +61 3 543-7795, Fax: +61 3 347-4783
----------------------------------------------------------------------------