From: Jim Peters (jim_at_uazu.net)
Date: 2002-02-13 10:19:57
Jim Meissner wrote:
> Please don't feel insulted! There is absolutely no insult meant.
I am not insulted, don't worry! Some things you say are obviously
really key pieces of information, and others I simply can't judge or
argue with without trying them out for myself. I read what you write,
and mostly I'm chuckling. I don't know how everyone else takes it,
but it's fine with me.
> I would love the option of looking at a portion of the file and play
> it, reverse and replay. How wonderful that would be!!!!!!!
I think this is the best way for me to put something in at the moment.
> How about starting with the linear frequency bins FFT like the Mind
> Mirror, Hal-4, ElectricGuru, etc. Then it would be compatible with
> existing display conventions. You can always add the log display
> later as an option. I also like the voice print display very much.
According to their web-site, the Mind Mirror III uses a series of
carefully matched filters, and they are not on a linear scale. The
frequencies are: 0.75Hz, 1.5Hz, 3Hz, 4.5Hz, 6Hz, 7.5Hz, 9Hz, 10.5Hz,
12.5Hz, 15Hz, 19Hz, 24Hz, 30Hz and 38Hz. So, not everyone uses FFT
and equally-spaced bins.
To me using an FFT and a linear scale is like going down a dead-end
street. Anyway, I already know how I'm going to do the analysis, so
taking a different approach to everyone else isn't going to slow me
down.
It will look something like the voice-print display, only the
frequency scale will be different on the left.
> Have you gotten the "Awakened Mind" by Maxwell Cade yet?
I haven't got an answer from the phone number on the web-site, and
I've been quoted 4-6 weeks from a bookseller because it has to be
ordered from the States ! I'll try again.
> I have read in some of your posts that you are concerned with the
> time delay of your filters to do biofeedback. Again, I believe you
> are listening to some information that I do not find to be credible.
Well, when I have a working EEG unit, then I can judge this. What I'm
really learning about here are the mathematical limits of the signal
processing techniques, nothing special about brainwaves.
There are different kinds of techniques that have different advantages
and disadvantages. These are the ones I've come across so far:
- Bank of IIR filters: small amount of work on each sample, provides
an immediate answer after each sample, unlimited flexibility in how
the bands are arranged, some blurring into the recent past.
- FFT of windowed data: medium-sized chunk of work every few samples,
limited to linear scales (equally-spaced bins), answers available
only every few samples, loses detail at low frequencies, too much
frequency detail and time-blurring at high frequencies.
- FFT-accelerated convolution of signal and a set of complex wavelets:
large chunk of work every few seconds, unlimited flexibility in
choosing frequency scale and amount of blurring.
This last one is what I plan to use. I've made it sound really
clever, but all it is really is the signal amplitude-modulated with a
carrier (to shift the interesting frequency to 0Hz), and low-pass
filtered with a Blackman window. In fact I'm combining the carrier
and the window to make something that looks like a wavelet, and the
FFT convolution is just a way to accelerate applying this to a large
run of data all in one go.
As you can see, because this is optimised for off-line analysis, I'm
getting farther and farther away from what is needed for real-time
responsiveness and biofeedback.
However, this will give some really good results and allow us to see
things buried in EEG recordings that we might never have seen before.
Jim
-- Jim Peters (_)/=\~/_(_) jim_at_uazu.net (_) /=\ ~/_ (_) Uazú (_) /=\ ~/_ (_) http:// B'ham, UK (_) ____ /=\ ____ ~/_ ____ (_) uazu.net
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:38 BST