discussion / Acoustics  / 26 March 2016

Acoustic monitoring for locating wolf packs

We currently use expensive Wildlife Acoustics SM3s to triangulate the position of howling wolves in Yellowstone National Park.  But we want to be able to deploy large numbers of wolf detectors based on more affordable solutions.  I'll be at the Science Makers meeting next week in Cambridge talking about my current system and ideas for developing smaller new detectors.  If anyone has ideas and will be nearby, please come along!

http://www.meetup.com/Cambridge-Synthetic-Biology-Meetup/events/227359402/




I will not be this meeting but would be interested in hearing ideas for making these detectors more affordable. I use Wildlife Acoustics SM3s for detecting tigers but am currently limited in how large of an area I can monitor due to the cost of the recorders.

What is the cost of the SM3? 

There are a lot of open source projects that could be modified/joined to develop an open source version of this technology.  As mentioned in the Raspberry Pi post, most of the code is written.  One would need to develop the source localization code.  Power management may be an issue.  Would solar be an option in the environments you are working in?  If not, some hardware level detector would need to be added in order to sleep the RPI and wake it up for processing. 

Take a look at 8SoundsUSB:

https://sourceforge.net/p/eightsoundsusb/wiki/Main_Page/

The device works with the standard linux image on the RPi.

This may be overkill, but some flavor of the above may work.

The same type of device could be used to track forest elephants and so on...

We had some good discussions on Saturday, and we will try to set up a dedicated discussion group.  Some of the main points were:

  1. The SM3 with GPS costs about $1000.  Our alternative should cost around $300.
  2. A single board computer is probably the best solution, although it's not clear which platform.  We will experiement with Raspberry Pi's for the moment, although Beagle Bones and even smartphone hardware were also mentioned.
  3. The ADC is going to be a key component, as the signals are usually very faint (I've attached an example).  The link you sent Eric is an interesting possibility for a beamforming application (measuring phase difference between nearby mics), but our current system uses multilateration based on time difference.  I suspect that the latter will be more accurate, but it would be interesting to try both.  In general, I'd like to see this as an open-source solution framework that can incorporate different technologies, approaches, and also of course, project-specific constraints.
  4. Power is an issue, in particular because of the data volume, which means that the download duty cycle will be very high.  Solar will be essential - I just hope it can supply enough to sustain the system, without a ridiculously obtrusive panel.
  5. The units must be time synchronised, which can either be via a GPS module for the SBC, or possibly via an RF receiver tuned to WWVB.  The former sounds much easier, the latter will probably use less power.

Looking forward to hearing what people think, and how we could make this a collaborative community project.

Arik

Hi, Arik, good to see you here.

I'm doing the localization on two scales in the last 1.5 years: right whale in Cape Cod Bay and Chickadees in our lab's backyard. Both have different physical scales but share a common challenge: a tiny time synchronization error will result in a huge localization error. I agree that Eric's suggestion on 8-sound is only good for beamforming method, which is less accurate than TDOA(Time Difference of Arrival)-based method IF the sound recorders are synchronized. That's a big IF to me! At least beamforming method rely on the microphones connected through cables and thus no time sync issue at all.

If you want to sync through RF or some other wireless protocol, based on my experience it might be very difficult to achieve it accurately since you introduce more syncrhonization error in the process. In addition, to reliably estimate TDOA, you need audio signals to be transmitted to be in a single computer. Transmitting an audio signal quickly drains your power.

I've done a simulation how the lack of time synchronization results in the localiation error. A 1 msec delay in one recorder is enough to screw things up quickly for a diamond-shape 4-unit array.

I'm pessimictic for a sparse array with wireless time-synchronziation in an affordable price range. For the baleen whale monitoring in our lab, we're replacing the original clock by atomic clocks, which are expensive but reliable no matter what environmental conditions are. For terristrial application, I tend to go with cabled systems and do beamforming to avoid the time-sync problem. It sounds technology backward but offers scientific location data of better quality. Scientific advance is what matters, isn't it?

Yu Shiu

Postdoc, Cornell Univ.