Thanks for all the very informative answers, I'll take a look (thanks
also for the kind words, Suzanne)
Sorry Steve, the links I posted were wrong (silly me!), these are ok:
http://groups.yahoo.com/group/naturerecordists/files/Member%20Files/20070616.blackbird.city.mp3
http://groups.yahoo.com/group/naturerecordists/files/Member%20Files/20070603.blackbird.pinewood.mp3
If anyone is interested, there are several other longer samples of
(urban) Turdus merula at my server, e.g.:
http://150.214.146.56:8181/_grabaciones/_campo/2007/20070616.trinos.04.mp3
Regards
D.
--- In Steve Pelikan <>
wrote:
>
> For some reason I can't download the sound files.
>
> However I've been interested in quantifying the similarity/differences
> between songs of several birds.
>
> One approach that has worked quite well for identifying different songs
> in a single individual's repertoire (see
> http://math.uc.edu/~pelikan/Dora/revi4/revi4.html for an example)
groups
> songs based on correlations between sonograms/spectrographs.
>
> A second approach selects specific elements from representative songs
> and estimates the similarities of these "probe" elements to the
elements
> that appear in each individual's song. Subsequent multivariant analysis
> (like clustering) serves to group the songs of different individuals
> into similarity classes. This has worked like a charm for Northern
> Cardinals.
>
> Finally, if you can decide on (at least as a hypothesis) certain
> specific parameters by which the songs may vary (lowest frequency
> present, rate of repetition of a certain element, etc.) you can measure
> these directly for each song. I did this with fairly good effect for
parids.
>
> If any of these approaches sound like they might be useful, you're
> welcome to software I've written to carry out the procedures.
>
> Good luck!
>
> Steve Pelikan
|