I don't know any one software package that does all of what you want (unless
you want to program in Matlab or something), but here's a way to do it by
combining two:
Use an automated detection system that outputs not only the time of each
detection, but also the strength of the signal there. Run it on both channels
independently, then take the resulting list of detections and open it as a
spreadsheet, with one line per detection. In the spreadsheet, calculate the
ratio between the signal strengths on the two channels, sort the list (the
entire spreadsheet) by this ratio, and then choose all lines (all detections)
whose ratio is above some threshold of your choosing. That should do it.
A plug for my own acoustic analysis software, Ishmael: You can do this task
using Ishmael. Use its "Energy sum" detector, and make sure the detector is
set to run on both channels and that it is writing the detections to a log
file. Run it on your data set and continue as above.... You still have the
problem that detections on the two channels are on successive lines of the
input file, so you'll need to check for this in the spreadsheet calculation.
Ishmael's automatic localization (bearing) estimator may come in handy for
calculating time offsets between channels. Feel free to email or call me
about
this if you like.
Ishmael is available for free at <a
href="http://cetus.pmel.noaa.gov/cgi-bin/MobySoft.pl";
rel="nofollow">http://cetus.pmel.noaa.gov/cgi-bin/MobySoft.pl</a>
.
Dave Mellinger
[EMAIL PROTECTED]
>From: Michael Noonan <[EMAIL PROTECTED]>
>To:
>Subject: seeking software advice
>
>I need a software recommendation. My goal is to automate a left-right
>detection process for stereo recordings.
>
>Background:
>
>I have a great many two-channel (stereo) recordings. They contain the
>underwater sounds of a marine mammal captive environment and they were
>recorded on a 24/7 basis over many months. Vocalizations coming from
>one side (left) are quite common on these recordings. Vocalizations
>from the right were rather rare, and we now have great interest in
>screening our tapes in order to detect these rare, right-sided events.
>
>I already have in place hardware and software for digitizing successive
>six-hour epochs as wav files and passing the large files via a local
>network to a high-speed processor for analysis. However, I have not had
>luck in finding a suitable software package for screening these files in
>the way that I need.
>
>Since these are stereo recordings from the same pool, any given
>vocalization appears on both the left and right channels. So, it will
>not be simply a matter of screening for events of a certain threshold on
>a single channel. It seems that I need software that can continuously
>compute a ratio between the left and right channel signal strengths and
>this ratio needs to feed into a triggering system (along with an
>ordinary intensity threshold feature) in writing out a time log of
>detected events. Alternatively, a program that was very sensitive to
>time-of-arrival differences between the two channels would presumably
>also prove useful.
>
>I would be grateful for any advice, and I will summarize responses for
>the list.
>
>Thanks in advance.
>
>Mike Noonan
>
>--
>
>Michael Noonan, PhD
>Professor, Biology and Psychology
>Canisius College
>Buffalo, New York 14208
>t: 716-888-2518
>e: [EMAIL PROTECTED]
>w: www.canisius.edu/~noonan/
>w: www.conservenature.org
>
|