naturerecordists
[Top] [All Lists]

Re: Audio editor/ DSP feature suggestions

Subject: Re: Audio editor/ DSP feature suggestions
From: Marty Michener <>
Date: Sun, 20 Oct 2002 16:25:14 -0400
At 11:40 AM 10/19/02 -0400, Steve Pelikan wrote:
>I'm putting together a progam that will perform audio editing and signal
>processing procedures and would appreciate any suggestions for features to
>be included --- especially features that would be useful for dealing with
>natural sounds. The goal at first is to provide procedures that aren't
>generally available on currently distributed programs. Since people use a
>variety of editors --- what are the most useful features of your favorite
>program?
>
>The program is writen mostly  in Java so that it shouldm run on ost
>platforms. It will be sufficiently modular that others can easily
>add/contribute code. I'd be particularly interested in hearing from someone
>who could contribute Java/C/C++ code to design (higher order) FIR/IIR
>filters.
>
>
>But ANY suggestions would be appreciated.
>
>Thanks,
>
>Steve Pelikan

Steve:

What a terrific project.  My compliments. Now that is really COOL!

As an early user of Cool (1996), who has recommended this product to 
hundreds of recordists, I have every year sent suggestions to the folks at 
Syntrillium that three new things were really needed to add to their editors.

I have NEVER received even a single acknowledgement for my email 
submissions, and now that Cool Edit Pro 2.0 is out, and incorporates none 
of these suggestions, I am boycotting their future products.  Ahhh,  for 
the days when you could email Dave Johnston, the inventor, and get a 
response in an hour.  Please note, I am cc.-ing them this posting, as 
well.  Maybe I can finally stir up something!

1. Look for sound piece "x" in file "y", or in all files in directory "\z"
This feature, which the original Canary implemented, is really a simple 
cross-correlation summation, as we did long-hand in the 1960's.  Canary 
took forever (15 minutes?) on old Apple Macs to look for a 2 second piece 
in a twenty second recording, but it was very useful for comparing some 
original sound you had recorded with published sounds, which you might 
suspect were your own stuff, used with or without permission.  It is sort 
of a "keep them honest" tool.  I recall the Cornell folks said it was very 
simple to implement.  Of course, with deliberate distortions, like 
time-stretching, or pitch shifting, correlation summed values are 
significantly disguised, but it is still of some use to us nature 
recordists, I think.

2. Cool always displays their spectra according to Kay Electronic's 
original format - frequency vertically, in linear format, and time 
horizontally, also linear.  We all on this group have been asking for a 
pseudo-semi-log vertical scale, more like a music scale, where each OCTAVE 
is granted so much vertical pixels, rather than a fully linear form where 
each kilohertz is granted a fixed unit.  Syntrillium has ignored this 
request, as far as I know.

3. Make the filter-construct screen oriented the same as the spectrum 
screen: frequency varying vertically.   Cool provides a very powerful 
editing band pass filter tool, including drawing your own complex filter 
curve, with clicks and drags.  Unfortunately, their frequency curve editing 
screen plots frequency HORIZONTALLY, inexplicably orthogonal to how it is 
displayed on the very file you are trying to edit!  I have been wrestling 
with this right-angle re-representation in my own work now for over six 
years, and it does not get much easier!

Cool does have what I say are three essential features to it, that you 
might try to 'preserve':

1. you can "preview" how it will sound, as you click and drag your curve, 
more or less, live; although it is really a pre-listen, NOT a 
preview.   This might be extended to a real preview, where the spectrum 
display would change visibly, as well.  This would be much more useful to 
me, actually.

2. you can set the frequency axis in this edit screen to either LINEAR or 
LOG, and go back and forth building the same curve as you go.  That is, the 
curve nodes you click-set in one mode, re-appear correctly represented in 
the other mode, and then back again.  This allows the nodes in the lower 
frequencies to be set accurately in LOG mode, then switch to LINEAR to 
adjust the higher bands more easily.

3. you can name and save to disk any curve or pair of morphed curves, and 
lastly

4. you can MORPH between any two such curves, and then designate a 
TRANSITION curve, that specifies, in the time domain, how the software 
switches from curve "A" over to curve "B", and back again, if you want.

I use this feature to its fullest.  Example: I often have a recording where 
unwanted bird "X" calls six times, and I make the transition curve to 
exactly remove each of the six sound intrusions, and I make curve "A" have 
the same number of nodes, but placed nearly flat in overall response, and 
curve "B" have the unwanted sound sharply filtered out.  When it is 
finished, the recording sounds like you just turned down the gain quickly 
on the unwanted sound at exactly the right times.

To restore some naturalness, I often select a part of the same recording 
with neither wanted nor unwanted species.  I copy it to a new file.  I then 
construct a filter manually that mirrors the removal filter ("B"), 
essentially generating a non-intrusive noise profile that complements the 
missing sound domain.  I then mix paste correct length pieces of this 
shaped noise into the holes left by the morph filter, leaving a recording 
that sounds fairly natural, of only the "wanted" species sounds, all at 
their natural level, timing and frequency levels.  [I am sure that Lang 
Elliott is shuddering, should he have read this far, but I hasten to add 
that MY purpose is only to instruct folks to learn about what to listen 
for, species by species, on my shoe-string budget, and is not to provide 
people with "natural" sounds, they might substitute for actually going out 
into the field.]

I realize how difficult it is to describe or understand this tool without 
seeing and hearing someone use it.

I have repeated posted to this list, how useful this tool is, and offered 
specific examples privately to numerous members, including exact 
screen-shots of how to use it, and so far, I think I am the only person to 
successfully employ it to actually "delete" specific noises, such as 
removing repeated blue-jay calls from a long recording of another bird.  I 
am afraid that with most users, the mind just boggles at all this.

The results are so spectacularly useful, I cannot believe I am the only 
person using Cool to its fullest, and would love to be proven wrong!  Maybe 
I/we can prepare a videotape ("over the shoulder").  It can be a kind of 
fun challenge.

Anyway, your "paint out a noise" would put this feature to shame, if it is 
able to incorporate the usefulness of the items already available.  You 
might seriously consider a "paint and replace" procedure, automatically 
performing what I do over a period of hours.  It would be necessary for the 
user to "paint" a quiet place where such non-intrusive noise could be 
sampled and copied.

I will cooperate in any way with your project.  I can provide before, 
during and after WAV files, etc.  Whatever you need.  My java is barely 
budding, rudimentary, though.


Marty Michener
MIST Software Associates
PO Box 269, Hollis, NH 03049


coming next week : EnjoyBirds - software that migrates with you.




________________________________________________________________________
________________________________________________________________________

<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU