I didn't mean to imply that machine recognition was impossible, just
impractical. The discussion began I believe with an ad for a CD (or DVD) ROM
purporting to have vocalizations of all (?) the species of birds (not frogs)
in ALL of North America including the Caribbean. The figure was put at some
2200 species if I remember correctly. Without getting into the business that
all the species of birds of this area have never been taped, that several
undescribed species exist within the area, and that some species are
essentially voiceless, even to get match-to-sample of a single vocalization
of each (and many have dozens of different vocalizations) would take multiple
life times. With all due respect to frogs most have a single highly
stereotyped song and even in Amazonia occur in about one-tenth the species
diversity as do birds in the same region.
Ted Parker first began censusing birds by tape recording them. In South
America birds in mixed species flocks may number 70-100 individuals of 30 or
40 species. These may move rapidly through the forest so that a visual
censuser sees only a tiny fraction of them. By taping the flock Ted could
later sit down and in real time listen and get at least a partial list of the
species diversity and density of a site. He introduced the technique to
Conservation International's Rapid Assessment (RAP) team efforts to identify
regions of the highest conservation priorities. Ted used to talk (wistfully)
about the day when computer programs might be able to take over some of this
tedious work. Unfortunately Ted did not live to see even the beginnings of
such an effort.
Today the value of voice recognition (still by humans, not by machine) is
actually voice non-recognition. There are field workers out there whose
knowledge of the vocalizations of birds and the biogeography of the continent
(in this case South America where my research interest lies) is such that
most as yet undescribed species are first being discovered because they hear
a voice that they instantly know MUST be new because they don't know it.
Usually they can put it to genus even before they see it, but some entirely
new genera are still be found whose vocalizations resemble those of no known
species. Additionally many near-mythical species known from only one or two
specimens from the 19th century are increasingly found to be relatively
common and widespread once the voice is recorded and recognized. Bioacoustics
is now an integral part of the alpha and beta taxonomic process in birds, but
I doubt there will ever be a "digital audio field guide to birds" though
future guides are sure to incorporate some of this technology. Anything is
better than "I want to see Miss Beecher"!
John Arvin
________________________________________________________________________
________________________________________________________________________
>From Tue Mar 8 18:22:42 2005
Message: 2
Date: Mon, 9 Sep 2002 08:32:55 -0700
From: Wild Sanctuary <>
Subject: Re: Match-to-sample software?
I got word back in March that there is a team of two fellas at Oxford
or Cambridge in the UK who have developed reasonable match-to-sample
software but have been unable to track them down. Anyone know who
that might be?
Bernie
> wrote:
>
>> This is an over simplified view of the process. The day when one can point a
>> microphone at a vocalizing bird and get a digital readout
>>identifying it will
>> never be a practical reality in my opinion.
>> John Arvin
>
>I doubt that will keep folks from trying. It's definitely a challenge.
>
>I've got a detector that I can point at any frog recording, and it will
>give me every species calling, even the faint ones! And with virtual
>100% accuracy. It can even do it out in the wild without a microphone.
>It's called John Jensen (of the Georgia DNR). Like it or not, that's the
>way we will be identifying what's on the recording, some person will
>have to do it. And some day I hope to be as good as John.
>
>There is some mild hope we can cut down on the empty space in
>recordings, however. Using some preset limits as to frequency etc. can help.
>
>Walt
>
>
>
>To unsubscribe from this group, send an email to:
>
>
>
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
--
Wild Sanctuary, Inc.
P. O. Box 536
Glen Ellen, CA 95442
707-996-6677 tel
707-996-0280 fax
http://www.wildsanctuary.com
________________________________________________________________________
________________________________________________________________________
|