Re: Excel 4.0, Fast Fourier HELP!!!

From
Date
Reply-To
>        Greetings,
>
>        This distress call goes out to anyone who has successfully used
>the Fast Fourier function of the PC-Excel 4.0 Analysis Add-on.  I've been
>trying to use it for shape analysis of cetothere and baleen whale ear
>bones, but the manual is useless, the Help files are useless, and the
>service reps at Microsoft don't know what to do with it.  Great.
>        Here's is my situation.  I have 49 near-mint (sorry, comic
>collecting term) bullae whose individual outlines have been digitized
>into sets of several hundred X-Y co-ordinates. According to Excel, these
>co-ordinates have to be entered within a single collumn in an X+Yi
>format.  It took several calls to Microsoft to discover that "i"
>represents an imaginary valve. Very helpful.  So the co-ordinates have
>been added and now I have several hundred single values.  I run this
>column through FF and receive several hundred X+Yi values with absolutely
>no explanation as to what they mean. Reading articles and texts involving
>Fast Fourier has not provided insight into how data should be inputed or
>how to understand the output.  It's times like this that make it
>difficult to argue in favor of PCs over Macs.
>        Again, I would love help from anyone who has experience with this
>program.  If you can suggest another Fast Fourier application, with very
>good documentation and a price that won't hurt a struggling grad student,
>I would be interested. Thank you.
>--
>        Matthew Paul Spizuco, Dept. of Geology, University of Pennsylvania
>        MSPIZUCO@SAS.UPENN.EDU

Matthew:

It seems that you are trying to use a open form (time series: series#,
value) FFT program to analyze shapes (closed form: position on shape,
vector from centroid).  I gather that you are trying to do shape
description/analysis of the ear bones.  You will need to change your x-y
coordinates to vectors (length from centroid of shape to edge of shape, and
the angle change as these vectors step around your shape).  What the
program will spit out when you feed in these values is a series of
amplitudes of harmonics which correspond to circularity (1st) harmonic,
elongation (2nd) harmonic, triangularity (3rd), "squareness" (4th), etc.
These amplitudes must be divided by the average radius in order to
normalize them.  The 1st harmonic is really an error term (it tells you how
well the centroid was located.

Once you've done all of your samples, the useful shape information in these
normalized amplitudes can be utilized by data reduction techniques (use
frequency distribution of amplitudes rather than the raw data).
Chi-square, maximum entropy, and other cool things are available.  You
should use the amplitudes which have the most shape-differential
information (largest range).

This is the limit of my knowledge.  Good luck.

A loyal Mac user.
--


--
Tim Demko       demkot@lamar.ColoState.EDU
Morrison Research Initiative
<URL: http://yuma.colostate.edu/~cwis70/morrison.html>
Department of Earth Resources
Colorado State University
Fort Collins, CO 80523
Office Phone: 970 491 6966
FAX: 970 491 6307