-----------
DESCRIPTION
-----------

EEG data was imported into an audio playing software at various new playback
sampling rates (greater than the original real sampling rate), as if it was
in PCM format (which of course it's not), and a voice recognition software
was used to interpret the resulting sounds, which didn't sound like voice,
but it did manage to interpret words out of it often.

It would be interesting to know what the subject having their EEG recorded,
was thinking, and their background (what would be in their mind) to see
whether there is any correlation between the found sentences/words and the
subject's thoughts, or whether it is in fact just random words/sentences
that are interpreted.

I suspect that it's probably just random. But I'll put up the found data
here anyway maybe it will be of use to someone. Admittedly I didn't apply
a 40Hz low pass filter like they said you needed to.

------
METHOD
------

1. Download Z.zip from
http://www.meb.uni-bonn.de/epileptologie/science/physik/eegdataold.html
and extract Z001.txt (EEG data with sampling rate 173.61Hz)
The reference for this data is "Phys. Rev. E, 64, 061907"

2. Modify Z001.txt by running a macro in crimson editor
(http://www.crimsoneditor.com/) that goes to the end of the line (hit END),
adds a comma (type ,) and then deletes the carriage return (press DEL),
save as macro 1 then hold down alt-1 so the macro runs continuously until
the whole file is converted. The result is Z001b.txt (It's a bit slow and
maybe there's a faster way, who knows).

3. Open Z001.txt in excel and make a line plot of the data in that column,
this is just for interest's sake, but not really required to do this step,
save it as Z001plot.xls

4. Copy the text in Z001b.txt and paste it into asc.htm instead of ...
where it says byteArr = [...], this gives asc.htm

5. Open asc.htm in firefox (or internet explorer but I used firefox),
do edit/select all then edit/copy. Then in crimson editor, file/new
and edit/paste and file/save as and save the file as Z001-asc.raw

6. With E-Speaking voice and speech recognition software
(http://www.e-speaking.com/download.htm)
Click on 'menu' then 'speech properties' then 'audio input', use this
audio input device (soundblaster should be selected or your sound card),
properties, use this audio input line, microphone, OK OK OK. Click on ON
next to menu so that the voice recognition software is active. Then say
"Start Dictation". Once dictation mode is activated (check by speaking
while the cursor is in a notepad window), go to speech properties, audio
input, use this audio input device, properties, use this audio input line,
and choose "what you hear". This will cause the voice recognition software
to take its input from what you hear.

7. Load a notepad blank file, and be ready to click into it when sound is
playing, so that the voice recognition software will type into notepad.

8. In the "audacity" software (http://audacity.sourceforge.net/) select
Project/Import Raw Data, Z001-asc.raw, then choose respective parameters
below (each of the sections below is a new import of the .raw file). For
example "200hz signed-8-bit-pcm big-endian" means signed-8-bit-pcm,
big-endian, and sampling rate of 200hz, for the import.

9. Press play on audacity and straight away after click into the notepad
window  click down the bottom of the window is fine).

10. Repeat step 9 a number of times, each time having the text appear on
a separate/new line of the notepad file, to distinguish the various things
it thinks the raw data file might be saying.

11. Save the file as an MP3 (eg 300hz signed-8-bit-pcm big-endian.mp3) so
that it's possible to play the file in future without using audacity.

12. Repeat steps 8 to 11 a number of times at different frequencies to see
the effect on changing the import frequency on what text is interpreted.

-----------
THE RESULTS
-----------

200hz signed-8-bit-pcm big-endian:
but.vote vote 
but.do 
it.V.

300hz signed-8-bit-pcm big-endian:
to both Goethe go because it did because been because because 
been because but to cut but 
because but to become because been become but 
it did J. but do but to go but 
it did Jude it did G. it did Joe Baca and do it could but 
it did chief but don't vote vicar to go and do but 
could do but vote vicar to go but 
could join beacon date vote 
could job back and back and back and do Baker can be 
could decode did June and dam beacon bacon bits and deter boat 
could decode gene do it vote to be convicted June beacon be 
could decode did June and don't beat can be can go to bed to cut but 
it did June beacon beacon to cut back its 
could don't do because bacon and don't be 
but 
it does take and take an eicon back at Baker can be 

400hz signed-8-bit-pcm big-endian:
aid good job but
Edu
and good big guy
but don't go back and
don't beat victim
bed
could good groomed 
could be converted big green beacon dead June 
don't become vacant bacon become because it can beat Joe 

400hz signed-8-bit-pcm no-endianness:
but 

500hz signed-8-bit-pcm big-endian:
obtained good guy 
could Grbac 
could go 
to bed good but groom 
beacon but can't been beaten 
bid by the can be can it go and 
bacon and it goes back 
button and 

600hz signed-8-bit-pcm big-endian:
Biden in Newton and then in
and vote in the in
and that and green
and good deeds
and and and
could gain
and 
could groomed
would and 
could and 
and bob Greene 
the need
mood and groomed 
and and good guide 

700hz signed-8-bit-pcm big-endian:
add good 
could good 
and and groomed 
and and and green 
and to continue to 

800hz signed-8-bit-pcm big-endian:
could and green 
condemned and green 
and and decline 
and decline 
and and bake and 
and and break and 
and and 
and and and and Dick and 
and and and and and 
and and and 
could and 

1600hz signed-8-bit-pcm big-endian:
poured
O.
O. O.

3200hz signed-8-bit-pcm big-endian:
R. R.

200hz 32-bit-float big-endian:
with

400hz 32-bit-float big-endian:
and gold 
could give 
and give 
the bid vote 
and 
did 
did bob 
Byrd did 
both did 
the bid 

600hz 32-bit-float big-endian:
and 

------
FUTURE
------

A future thing someone could do, is to measure EEG while the subject is
listening to mono music though headphones, and record the music alongside
the EEG so that correlations can be made to see whether the EEG contains
the music (or some portion of its frequencies) encoded somehow in it. Also
this experiment could be done again, this time with the subject thinking
specific words in their mind, to see if any of those words come up in the
EEG derivation.