Forgive me if this has been mentioned before...
Yesterday I attended a conference by the Australasian Sound Recordists
Association (ASRA) on the topic of 'Endangered Sounds'.
An engineer by the name of Neil J Boucher presented a fascinating demonstration
of software that is able to identify species by the sounds they make.
Apparently it has similar accuracy to human listeners but works considerably
faster.
The applications and possibilities offered by this software are quite
interesting.
It is currently being used in autonomous recording stations placed around
Australia to log the calls of endangered species. Known calls of the particular
species are loaded into the software. The system then continually monitors the
sounds captured by the microphones and, when it detects a match, records it.
This makes a very intelligent form of remote recording that only records what
it is looking for, rather than recording everything. It saves gigabytes of hard
disk space and hours of listening/logging time.
It is also being used to capture the sounds of species for which there are no
known recordings. In this application, a number of sample recordings are made
of a particular habitat where the species is believed to exist. From these
recordings, all known sounds are logged and loaded into the software. The
system then monitors the microphone signal and only records sounds it *doesn't*
recognise, thereby acting as a filter.
Perhaps I am naive, but the technology and the demonstration were quite
impressive. As were the autonomous recording stations he has built, which allow
recording for months at a time. More information here:
http://www.soundid.net/
The 'Research Papers' link has some particularly interesting stuff...
- Greg Simmons
|