Automated processing- I have a very large amount of data to get through!

[Erin Westerhuis]

Hi Everyone, 

My name is Erin Westerhuis and I am a phd student based in Alice Springs looking at seasonal patterns of bat activity in ephemeral river corridors. I'm using Anabat Express for recording and I've started manually identifying calls in AnalookW (based on call libraries from elsewhere and also from my own small collection of reference calls) but it's taking a very long time, some of my summer nights have >4000 recordings and I have 48 detector nights per season with six seasons so far. At this rate I will be finishing my phd in 2028!

To try and filter out the partial calls and insects, I've tried building filters for all bats which is ok, then a spearate one for each species or group but that seems a bit hit and miss ie when I go back and check the calls that didn't pass, some of them appear to be fine.

I've read about the use of Anascheme in other projects but don't fully understand if that's something that I can use and whether it would help me. 

If there a way to group calls that are similar and then identify them post hoc? Sort of like unsupervised clustering

Any help most appreciated 

Cheers,

Erin

[Jo]

Hi Erin,

I created a random forest model to automate species ID when I was going through loads of WA bat calls for my PhD. I can send you my R code for you to create your own regional specific model if you'd like, as well as the documentation/instructions for getting the parameters out of AnaLook and into R. You'll just need a reference library of calls for your area to create a region specific model.

Cheers,

Jo

[Julie Broken-Brow]

Hi Erin,

Titley Scientific has a new software program Anabat Insight, that may help you with batch processing. You can tweak your filters in real time and runs searches that automatically label your files based on your defined parameters. We also have a decision tree tool which you can use to build your own ID system. If any other this is of interest to you, just shoot me an email and I can go into more depth (juli...@titley-scientific.com) . At the very least it can help you quickly process the noise files out.

[kyle armstrong]

Hi Erin, just a quick reply from me as I am traveling. In your corner of the world, the analook insight filter and decision tree system would be a very appropriate choice. Many years of thought and development have gone into the analook system. I am using that for a large project on south Australia. Jo"s kind offer of her random forest walks r code is also a good opportunity to get into that type of analysis. Her code works very well.  Good luck with it, and ask questions if you get stuck for too long. Kyle

On 16/01/2018 6:31 am, "Julie Broken-Brow" <nocturna...@gmail.com> wrote:

Hi Erin,

Titley Scientific has a new software program Anabat Insight, that may help you with batch processing. You can tweak your filters in real time and runs searches that automatically label your files based on your defined parameters. We also have a decision tree tool which you can use to build your own ID system. If any other this is of interest to you, just shoot me an email and I can go into more depth (julie.bb@titley-scientific.com) . At the very least it can help you quickly process the noise files out.

On Monday, January 15, 2018 at 11:23:54 AM UTC+10, Erin Westerhuis wrote:

Hi Everyone, 

My name is Erin Westerhuis and I am a phd student based in Alice Springs looking at seasonal patterns of bat activity in ephemeral river corridors. I'm using Anabat Express for recording and I've started manually identifying calls in AnalookW (based on call libraries from elsewhere and also from my own small collection of reference calls) but it's taking a very long time, some of my summer nights have >4000 recordings and I have 48 detector nights per season with six seasons so far. At this rate I will be finishing my phd in 2028!

To try and filter out the partial calls and insects, I've tried building filters for all bats which is ok, then a spearate one for each species or group but that seems a bit hit and miss ie when I go back and check the calls that didn't pass, some of them appear to be fine.

I've read about the use of Anascheme in other projects but don't fully understand if that's something that I can use and whether it would help me. 

If there a way to group calls that are similar and then identify them post hoc? Sort of like unsupervised clustering

Any help most appreciated 

Cheers,

Erin

--
You received this message because you are subscribed to the Google Groups "Australasian bat call group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kilohertz+unsubscribe@googlegroups.com.
To post to this group, send email to kilo...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Erin Westerhuis]

Thanks Jo, that would be amazing. My email is erin.we...@cdu.edu.au 

[Erin Westerhuis]

Thanks Julie, I'll email you :) 

[Erin Westerhuis]

ok thanks, yes I was getting a bit overwhelmed with doing it manually!

On Tuesday, January 16, 2018 at 11:57:12 AM UTC+9:30, kyle armstrong wrote:

Hi Erin, just a quick reply from me as I am traveling. In your corner of the world, the analook insight filter and decision tree system would be a very appropriate choice. Many years of thought and development have gone into the analook system. I am using that for a large project on south Australia. Jo"s kind offer of her random forest walks r code is also a good opportunity to get into that type of analysis. Her code works very well.  Good luck with it, and ask questions if you get stuck for too long. Kyle

On 16/01/2018 6:31 am, "Julie Broken-Brow" <nocturna...@gmail.com> wrote:

Hi Erin,

Titley Scientific has a new software program Anabat Insight, that may help you with batch processing. You can tweak your filters in real time and runs searches that automatically label your files based on your defined parameters. We also have a decision tree tool which you can use to build your own ID system. If any other this is of interest to you, just shoot me an email and I can go into more depth (juli...@titley-scientific.com) . At the very least it can help you quickly process the noise files out.

On Monday, January 15, 2018 at 11:23:54 AM UTC+10, Erin Westerhuis wrote:

Hi Everyone, 

My name is Erin Westerhuis and I am a phd student based in Alice Springs looking at seasonal patterns of bat activity in ephemeral river corridors. I'm using Anabat Express for recording and I've started manually identifying calls in AnalookW (based on call libraries from elsewhere and also from my own small collection of reference calls) but it's taking a very long time, some of my summer nights have >4000 recordings and I have 48 detector nights per season with six seasons so far. At this rate I will be finishing my phd in 2028!

To try and filter out the partial calls and insects, I've tried building filters for all bats which is ok, then a spearate one for each species or group but that seems a bit hit and miss ie when I go back and check the calls that didn't pass, some of them appear to be fine.

I've read about the use of Anascheme in other projects but don't fully understand if that's something that I can use and whether it would help me. 

If there a way to group calls that are similar and then identify them post hoc? Sort of like unsupervised clustering

Any help most appreciated 

Cheers,

Erin

--
You received this message because you are subscribed to the Google Groups "Australasian bat call group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to kilohertz+...@googlegroups.com.