naturerecordists
[Top] [All Lists]

Re: Normalizing calls in AviSoft

Subject: Re: Normalizing calls in AviSoft
From: "Raimund" animalsounds
Date: Mon Jan 11, 2010 7:10 am ((PST))
alizaleroux wrote:

> What I'm trying to do is extract individual calls that occur in a train o=
f 3-6 calls. I need to put these calls into separate wav files yet measure =
their original intercall intervals accurately in the original file. I don't=
 want my interval measures to be affected by the differences in overall cal=
l volumes. If I have to do it all individually some calls might appear shor=
ter/ longer simply due to overall amplitude. Due to background noise I inte=
nd labeling each call by hand so as to not confuse Avisoft's automatic meas=
ures.

Hi Aliza,

I think that you probably don't need to put the individual calls into separ=
ate wav files. The Automatic Parameter Measurements tool in Avisoft-SASLab =
Pro can already handle varying call amplitudes. You just had to use the "El=
ement separation" option "automatic (two thresholds)" that employs an addit=
ional threshold (relative to the peak amplitude of each call) for locating =
the start end end of each call. However, if the amplitude differences were =
too high, it might indeed be required to treat the individual calls separat=
ely. If you are unsure, you could send me (to support <at> avisoft.com) one=
 of your recordings as an email attachment. I would then try to figure out =
the best settings for your specific application.


> About CORRELATOR -- are the comparisons not affected somewhat by amplitud=
e? I realize that the program examines the spectogram as a whole, but it fe=
els like calls should be at least within a small amplitude range instead of=
 all over the place... Is this true, or can I compare soft and loud spectro=
grams (of the same call type) without fear? I want to look at certain indiv=
idual call parameters but running the correlator in addition to this certai=
nly can't hurt.

The CORRELATOR does not care much about different spectrogram amplitudes be=
cause there is already an inbuilt normalization procedure that takes care o=
f that issue.

Regards,
Raimund








<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