Using multiple AudioMoths for sound localisation

Hi Paul,

I am in the midst of fieldwork so slow in reply. I still use homemade routines for localization. Works for me but i am afraid it's too losely knit for someone else to work with it right away. For something more thorough i would advice to check out Opensoundscape, that package seems to include localization. What project are you working on?

Cheers,

Eric

Hi Alex,

This is not more that a first impression but I am very positive about it as well. I realized i could improve by adjusting for time-drift per instrument and the difference in sample-rate connected to that. However the drift as shown in the (very handy) GPS logfile was comparable for the three devices although not exactly the same off course. The distances in the case of Bittern are much greater that what they would be in the case of (e.g.) songbirds on a smaller patch and therefor the relative error is smaller and the booming area of interest larger which makes the whole effort easier .

I am quite certain that for next season there will be a grit of synchronized Audiomoth's in operation here over a considerable (N2000) area of interest which will be of real conversational importance i reckon.

Greetings,

Eric

Sounds really promising. Correcting for the HFXO is quite straightforward as you can just adjust the time measurements by the fact that the sample rate wasn't exactly the same on all devices so the actual time is offset slightly from the start time. In our experiments this gave us about a 20ms offset over an hour which corresponds to a position error of about 5m. The LFXO offset is likely to be smaller depending on the gap between setting the time and starting the recording. Getting to a few milliseconds time accuracy, and probably less than 1m error, is probably achievable. Although our simulated gunshot was an impulsive sound so it was quite easy to pick the start time of the event. This part might be the most uncertain bit in other applications. We have some other firmware which will align the recordings to about 1us by keeping the GPS running continuously and recording the number of samples as each pulse per second (PPS) event arrives from the GPS module. We've done lab tests with this version. However, I think you probably need something that produces a frequency sweep to really be able to correlate signals with sufficient accuracy to get the true benefit of the better time synchronisation.

Hi Alex,

No i just used the times measured as such (for Bittern i use the time of peakfrequncy which is determined by Ravenpro, since this is so well defined in this species and this probably is not affected by the traveling of the sound over distance so the error of (human) measuring is minimized), i did use multiple short recording periods so the GPS synchronization moment was never far from my actual recordings. We are still in the progress of finishing our program so hat it will give us coordinates as a result of the estimate bases on entering the coordinates of AM's plus time measurements for each recorder. For this test i used a visual approach that i can email to you if you wish.

Cheers,

Eric

Hi Everyone,

I have my first results on localizing Great Bitterns. The test was done in an area where i know the exact location of booming-posts. I used a triangle of the Audiomoth/GPS combinations placed about 400 meters apart. Measuring the exact time of arrival on the spectograms and a program of our own making gave me three hyperbole's crossing where the sound should come from. The results are as good as spot on! For one bird, that booms on a distance of about 2 km's from the center of my triangle the estimate of its location was perfect in terms of direction, the exact distance is (logically) more difficult to establish in such a case.

Greetings,

Eric

Hello Team, Andy,

That is very good news! Thanks for the effort you put into this most valuable extension. For my plan of testing the Time-Of-Arrival-positioning of Bitterns that would be well within the deadline. If i could start say the the first half of March all would go according to plan.

If it would be helpful me doing some more (field) testing earlier before the season we could contact about using the prototypes, i will reply to your email-address Andy so you will have mine as well.

Cheers,

Eric

Hi Eric, We are hoping to make these available by the end of February in a GroupGets campaign. If you require them earlier we might have a few prototype available, email andy@openacousticdevices.info to find out more.

Hi Alex,

Is there a chance that the GPS modules will be available before the start of this years breeding season?

@David Jarrett, i have send you an email with a link to a page where you can experiment with triangulation, did that arrive? Anyway I am working on an alternative method for synchronization and will post the results here if these would be worthwhile.

Kind regards all,

Eric

Eric - what is the programming package/solution you are going to use for the localisation of Bitterns? I'm hoping to do a similar thing with Curlew territories Alex - I'll be really keen to join the Groupgets for the GPS boards as well, that degree of accuracy would be great. Cheers,

David

Hi Alex,

That really is great news! Count me in for a group purchase, if that is possible that is. There will be an anouncement of this i expect and will look for it every day! This level of synchronisation would certainly do it for localisation of Bitterns. Can't wait for next season.

Many thanks,

Eric

Hello, any news on the status of this GPS synchronisation solution? I did a very succesfull project on Bitterns using Audiomoths and already collected a ton of useful data (thanks for making this open source project happening! It realy made a difference). If i could get the timekeeping accurate for next season it would be straightforward to add sufficiently accurate localisation of the booming males. Programming and math to do this allready in place so hopefully I will find a solution to sufficiently synchronize the 8 Audiomoths. One sulution would be to measure the drift for every particular divice and use a correction when analyzing. Another is to make a simple device that produces a accurately timed beep and put this 'on tape' when i place a Audiomoth in the field and again when i collect it. Any other suggestions maybe.

Cheers,

Eric

Hi David, It depends on how accurate you need the synchronisation to be compared to the relative time difference between events. When you set the time (either with the smartphone app or the desktop app) the time within each recording will initially be accurate to +/- 10 ms. However, the clock will drift relatively quickly - up to about 1 second per day. We have a GPS synchronisation solution that we've been experimenting with which will maintain at least 20us synchronisation between recordings, and we should be able to release some hardware and firmware for this later in the year. Alex