Re: SETI public: Pattern Recognition.....

From: David Woolley (david_at_djwhome.demon.co.uk)
Date: Sun Jul 06 2003 - 14:03:13 PDT

  • Next message: LARRY KLAES: "SETI public: Fw: [DarlingsSpace] David Darling's Newsletter #13"

    > > > But even in this, there needs to be some controls, such as ignition noise

    I believe, ignition noise tends to be rejected by the FFT if you use
    the mean power across the spectrum as reference.

    > > > is of no interest to me as are wide band power supply birdies etc.

    Narrow band birdies are eliminated by cataloguing them. More difficult,
    for an amateur, is cataloguing non-local sources, but professionals
    do exactly that. Wideband signals are rejected by low pass filtering the
    output of the FFT, e.g. using adjacent channels as a reference level.

    With sub-1Hz bandwidths used by professionals, static sources can be
    rejected by the lack of a chirp in the frequency (which results from
    the acceleration due to the earth's rotation and orbital motion).

    In terms of recognizing patterns in individual FFTs, amateurs generally
    integrate across the whole beamwidth. I think the most common technique
    is to do that visually, by looking for steaks in a waterfall display.
    Chirp is visible at this level.

    Doing it visually means that they look at a whole day's results at once.

    However, the only really reliable way of confirming a non-local source
    is to take measurements from geographically diverse locations, and small
    dish SETI actually needs to do that very quickly, otherwise the much
    greater range of big dish SETI means that big dish is going to be much
    more likely to find a signal than even enough small dishes to give
    continuous full sky coverage (the disadvantage of large dish SETI is
    that it can only look at any one direction for a very small amount of
    time, but will eventually find a signal that is present for years; only
    when you get to signals that last for less than an hour does the
    unlikelihood of pointing in the right direction begin to dominate the
    advantage of greater range).

    > I have since your first email put some hours into SETI@Home looking for
    > information on the fast folding algorithm. I have not found anything, so
    > your help could be very valuable in finding out what has already gone one.

    <http://groups.google.com/groups?selm=39944AB0.7F6C0E0A%40setiathome.berkeley.edu>

    The other specific test that SETI@Home uses is to match the beam shape
    at medium frequency resolutions (high resolutions don't have adequate
    time resolution). Because the pattern can be approximated by a gaussian
    and these occur very commonly in statistics, I suspect there is a lot of
    code around for doing this, but I haven't researched it.

    >
    > If you know of any general texts on this subject, that are detailed enough
    > that one can implement them with out Post Graduate math skills I would
    > appreciate that as well.

    I doubt that there is any postgraduate maths involved, although I don't
    have access to an extensive maths library to research what is avaliable.
    I haven't got round to reading the source code yet.


  • Next message: LARRY KLAES: "SETI public: Fw: [DarlingsSpace] David Darling's Newsletter #13"

    This archive was generated by hypermail 2.1.6 : Sun Jul 06 2003 - 14:22:11 PDT