From dwbrown@nps.navy.mil Tue Jul  5 17:19:51 EDT 1994
Article: 2782 of comp.speech
Xref: glinda.oz.cs.cmu.edu comp.soft-sys.matlab:7057 comp.dsp:11910 comp.speech:2782
Path: honeydew.srv.cs.cmu.edu!nntp.club.cc.cmu.edu!newsfeed.pitt.edu!gatech!news-feed-1.peachnet.edu!emory!swrinde!elroy.jpl.nasa.gov!lll-winken.llnl.gov!taurus.cs.nps.navy.mil!nps.navy.mil!usenet
From: dwbrown@nps.navy.mil (Dennis W. Brown)
Newsgroups: comp.soft-sys.matlab,comp.dsp,comp.speech
Subject: SPC Toolbox
Date: 30 Jun 1994 23:06:44 GMT
Organization: Naval Postgraduate School, Monterey, CA
Lines: 81
Distribution: world
Message-ID: <2uvj64$db@nps.navy.mil>
Reply-To: dwbrown@nps.navy.mil
NNTP-Posting-Host: in50210.cc.nps.navy.mil

To all:

I've place updated version of spctools.tar.Z and spctools.zip at

ftp.nps.navy.mil:/pub/ece/spctools/spctools.tar.Z
ftp.nps.navy.mil:/pub/ece/spctools/spctools.zip

as of 1500, CA time, Jun 30th.  This would be the last version but
as I'm still not done with all the documentation (approaching 300+
pages), there will be one last post next week.  I'm currently living
out of a suitcase and pickup, just hanging around Monterey to finish
this thing.

Major changes in this version.

1.  I've "speeded up" ZOOMTOOL.  When you grab the cursors by clicking
on the zoomed line (not the cursor itself), the position readouts are 
updated continuously vice only when you release the mouse button (during
drags).  Doing this required me to re-write much of the zoomtool code.
To get the maximum speed, I converted it all to the UserData programming
paradigm and use the Global Varible paradigm during cursor drags.  The 
speedup occurs by eliminating some of the problems caused by having to
use the UserData (or, in this case, the line objects XData property) 
for variable storage.  For example, to update the Y positional readout, 
I have to find the Y value for the current X value from the line's 
YData property.  This first requires "getting" the YData with a call
sorta like "y = get(h,'YData')".   The trouble with this and the whole
ideal of using UserData properties for variable storage is that even though
I need just a single value out of the vector stored in YData, I have to
first make a copy of the whole YData vector.  Once that is accomplished,
I can then use subscript notation to retrieve the Y value I need 
"yval = y(index)".  For most of the buttons ('<', '>', etc), I still have to 
do this but when dragging, I make the copy once on the buttondownfcn call
and stores it as a global variable.  The 'windowbuttonmotionfcn' and
'windowbuttonupfcn' just access this variable.

The MathWorks needs to
provide a better way of storing variables for GUI applications use. 
Having to retrieve entire vectors and matrices for single values requires
way too much overhead having to first copy the whole damn thing.  I've
made a few suggestions to them on this but haven't heard anything back.
At minimum, what is needed is an "Variable" object that can be a child 
of any other object.  The variable object would have at least one property
which is the data itself (call it say 'Data').  It could have additional
properties such as Maximum, Minimum, Mean, Size, etc, which would speed up
retrieving these values (e.g. they are set when Data is set, thus the 
corresponding functions won't have to be executed on them).  One could
the retrieve these by "m = get(h,'Maximum')" where h is a handle to the 
object.  Note a Tag property would also be included so that the handle to
a variable can be found with "h = findobl('Tag','myvariable').  Individual
cells in vectors and matrices could be retrieved ( or blocks for that
matter) with call to "y = get(h,'Value',index)" for single vectors,
"y = get(h,'Value',[x y] for matrices, and "y = get(h,'Value',start:end)"
and "y = get(h,'Value',[xstart xend ystart yend])" for blocks.

Best thing though would be to simply give each figure window it's own
workspace with persistnent storage so you don't have to worry about
retrieving variables from any objects.

2.  I've added a function called "spcscope" which is used like the
rest of the GUI functions.  Try calling it with "spcscope('seatsit.voc')".
It provides a way to playback signal looking at the time-domain and
spectrum of the signal.  It makes it look like you're sitting in front
of both an o-scope and a spectrum analyzer.  The data is framed to 
the time window you specify and successive frames can be overlapped.
Different smoothing windows are available and you can smooth the 
spectrum in the frequency domain (averaging adjacent frequency bins)
or time domain (averaging successive frames after computing the FFT).
This is still a work in progress.

Regards,
Dennis


---
| Dennis W. Brown           |
| Naval Postgraduate School | email: browndw@ece.nps.navy.mil 
| Monterey, CA 93940        | Usenet: dwbrown@cc.nps.navy.mil
| (408)656-2393             |    CIS: 75450,1105