Computers Seek The Call Of An Extinct Bird

Buran writes: "As a self-proclaimed geek and a relative (and fascinated!) newcomer to the world of birding, I found this article in the New York Times Science Tuesday about the search for the ivory-billed woodpecker to be rather interesting. The bird, which was listed as extinct in 1997, has not been definitively sighted since the 1950s, but a recent reported sighting (in 1999) has led to a redoubled effort to find it. The geek side is this: Since it would be impractical for a human to sift through 5,000 hours of recorded sound (two and a half years, they estmate) to listen for the bird's distinctive call, the Cornell researchers are working on algorithms that can pick out interesting sections of digitally recorded sound, taken from microphones placed throughout the study area, for a human (who can outdo a computer any day at making the final determination) to review. I am hopeful that the search will return a positive result."

Humans better?

[Deanasc]

I would think that the computer can match exact wave forms far better than the computer could.

Re:Humans better?

[spt]

If it is a long complicated call, interspersed with other bird noises, the computer can find the frequencies but the human is better at the pattern recognition to see if it is the right call or a freak concurrence of other noises with a similar frequency signature.

Re:Humans better?

[guiding_knight]

Exact waveforms yes, but when was the last time you heard two identical bird calls, even by the same species, or the same bird? An exact waveform match just will not be flexible enough for this application, thus the general match and a human reviewer.

Re:Humans better?

[Ledge]

You would have to assume that all ivory billed woodpeckers sounded exactly alike, which they don't. The human brain is way better at fuzzy stuff than a box is. I once saw a program on TV about this in regards to facial recognition. The computer was great at it, as long as it was easy to get the primary facial measurements. Throw in some distortion and the computer got lost on a sample that the human eye could easily identify

Re:Humans better?

[drewbradford]

There is no doubt that the human brrain is better at the "fuzzy stuff" than a computer is, but computers aren't always as bad at it as some people think.

I know that there are a lot of "intelligent" matching algorithms now being developed and deployed that can see patterns even with distortion. I have heard rumors of some facial recognition software that was being considered for deployal at O'Hare International Airport last year (I don't know if it actually was or not), but the technology seemed to be proficient at picking up criminals/suspects even if they did some altering of their facial features, such as grow a beard, wear thick glasses, or get a tan.

My point is that certainly the human brain is better at picking up the "fuzzy stuff", but computers are getting better and better at it.

Re:Humans better?

[bmorton]

I would think that the computer could match exact wave forms exactly the same as the computer could.

Re:Humans better?

[panthro]

Only as long as it's not using Windows.

Hrm indeed.

[DocMiata]

And after sifting through 5,000 hours of sound, they discovered that what they heard was the kid down the street playing Sim-Ivory-Billed-Woodpecker 2002.

Solution...

[wholesomegrits]

The geek side is this: Since it would be impractical for a human to sift through 5,000 hours of recorded sound (two and a half years, they estmate)

This is a trivial problem: you pay the listener by the hour.

RealTime plus Camera?

[n2kra]

can't read the article (registration) BUT

I wonder if it can be done RealTime, (how good are the current gen DSPs) and the
direction triangulated with 3 or more mics to point a camera in the right direction?

Re:RealTime plus Camera?

[Hektor_Troy]

And once the bird has been located, we cover the area with poison, to get the bird collected, just to prove, that it's not extinct.

Seems so simple

[Zelet]

This is a simple observation, but it still amazes me that as far as we have come with computing power, we still haven't even come close to the fuzzy abilities of nature/evolution. Even raw computational power of the human brain is staggering. It makes me humbled when I think of how much better nature is at this "programming/hardware" design than we are! :)

Re:Seems so simple

[ShavenYak]

It makes me humbled when I think of how much better nature is at this "programming/hardware" design than we are! :)

Of course, keep in mind that Nature has been building brains for millions of years; we've only been doing it for about fifty.

But what shall we name it?

[Traal]

PECK@home?

These are great projects

[bstrahm]

Having worked to identify the spread of Oak Wilt using digital imaging in the mid 80's, I love these biology projects. These problems aren't as easy as you think, given that you don't REALLY know what the bird sounds like (a 50 year old recording maybe ???) you don't know what else is happening in the area (did that jet just fly over ???) how the mic is tuned, etc. etc. etc.

I would love to sit in on this project and try to figure out what the various sounds are...

Ivory billed.......ught oh......

[CDWert]

These things have an ivory colored bill....and make a wierd sound ?

Ooooppppssss.....

I think I hit one last year, thought it was just a mutant, seriously. Ive seen woodpeckers around the house in droves since I was a kid, there a lot of older poles riddled with bugs they like to chomp, It can get at times annoying, no worse than catching a humminbird in the eye, had that happen too.

But whizzing down my moms road last year, smack right in the windshield, I pulled in my moms and walked back, it was dead and I wanted to make sure it wasnt suffering. My uncle, next door commented it was indeed different than we had ever seen, the usuall red and grey jobbers.

Chock one more extinct speices up to GM.....

NPR story has original recording

[Cy Guy]

NPR has done two stories on this in the last couple the months. The first was a piece specifically about the 1935 audio recording [npr.org]. And the second talks about this latest attempt to locate any surviving members of the species [npr.org].

Could the human parallel process the audio?

[esnible]

If the human listened to, say, 10 audio streams at once, wouldn't he or she still be able to hear the distinctive cry of the woodpecker?