Seen in Wired News:
Link:
http://tinyurl.com/k7vjx
Text:
Parsing the Puffin's Patois
By Rachel Metz
02:00 AM Jul, 27, 2006
Mr. Ed spoke his mind in plain English, but most animals are not so
easy to understand. Now, a group of scientists is trying to decode the
meaning behind the growls, clucks and neighs, hoping to improve
captive animals' lives and humans' abilities to track their numbers in
the wild.
Called the Dr. Dolittle Project, the effort includes scientists from
several universities and institutions. They are studying animals and
building a database of the sounds they make, along with accompanying
behaviors.
"Our basic goal is to understand what they're saying, at some level or
another," said project leader and Marquette University assistant
professor Mike Johnson.
To do this, researchers are applying human speech technology to animal
chatter. For the past five years, Johnson and his colleagues have been
recording animals like African elephants, dogs, whales and chickens,
and videotaping or observing the concurrent behaviors, such as
swimming or eating.
Researchers feed observations and sound recordings into a computer
that uses algorithms to analyze and group the sounds according to what
the scientists would like to study (for example, calls elephants make
while playing). Each sound can also be matched with an individual
label, marking things like the animal it came from.
Now, researchers are adding more sounds to the computer database
without telling the computer associated behaviors or which animals
they come from. By doing this, they hope the computer can learn to
determine what the sounds mean, said Pete Scheifele, an assistant
professor in residence at the University of Connecticut and one of
Johnson's Dr. Dolittle Project colleagues.
In developing a computer-based speech-processing technology with
remote monitoring systems, the researchers imagine it could be used to
alert aquaria managers or zookeepers when their animals are
distressed, or keep tabs on species' populations in the wild through
tracking their unique calls. Or it could help ranchers or zookeepers
recognize changes in their animals' moods, assisting them to keep the
beasts content or make appropriate adjustments.
"We're not up to that point yet, but from an animal welfare point of
view that's one of the things this computer model will let us do,"
Scheifele said.
The speech-recognition technologies could also help researchers
develop methods for conducting a census on animals in the wild.
Eventually, they hope to set up a wireless microphone network
somewhere, to record animal sounds and transmit them to a distant
computer where an algorithm would analyze them and give an estimate of
the number of a certain species in the recorded area. This would be
much less invasive than species-tracking methods like capturing,
tagging and then re-spotting animals, and easier when dealing with
flighty animals, like birds, Johnson said.
"I think it's possible," Johnson said, though he added, "I think there
are quite a few technical hurdles to overcome before you could get to
a deployed system." These include a cost barrier; they'd want to use
dozens of microphone nodes, and he estimated each would cost several
hundred dollars.
Douglas Nelson, director of Cornell University's Borror Laboratory of
Acoustics, which studies and archives animal sounds, said
automatically recognizing animal sounds can be difficult because they
can be variable.
"Just discriminating animal sounds from background noise can be a
considerable challenge," he said.
Background noise is definitely a concern to Dr. Dolittle Project
contributors -- some of Scheifele's work centers on the effects
ambient noise has on animals, and how it could impede their ability to
chat with one another.
Over time, researchers want to build their sound library with more
animals, and more exotic animals like dolphins, tigers and rhinos.
Scheifele is already recording horses with University of Rhode Island
adjunct professor David Browning, as part of what they call the Equine
Vocalization Project; this data should eventually be tapped by the Dr.
Dolittle Project.
Could all this lead to a better version of 2002's doggie translator,
the Bowlingual?
Sure, but it's not just a matter of creating better computer
algorithms to analyze creature calls, Johnson said. "You need a better
understanding of the animal communication process," he said.
|