search

WavX / bioacoustics

bioacoustics package: detect and extract automatically acoustic features in audio recordings
44 stars 13 forks source link

blob_detection and threshold_detection give very different results #15

Open AugustT opened 3 years ago

AugustT commented 3 years ago

While these two functions usually isolate the same calls, the metrics returned are quite different. This can be seen using the example data shipped with the package. I'm interested in the peak frequency/frequency centroid values, and while from the plots it is clear both functions have picked out the calls the metrics produced are very different. Do you know what is going on here/

fabianekfrancois commented 3 years ago

Not at all...

...just kidding!

The main reason is that blob_detection and threshold_detection are indeed two very different functions. To put it simply, they do not rely on the same algorithms to detect and extract call parameters. More info on that is given in the package index files: https://cran.r-project.org/web/packages/bioacoustics/index.html and more specifically the introduction page, where both functions are described in more details: https://cran.r-project.org/web/packages/bioacoustics/vignettes/introduction.html

For your purpose I would recommend using the threshold_detection function since it relies on amplitude pic detection instead of groups of pixels "entangled together" (See Chang et al., 2004). Two extracted call parameters from the threshold_detection function might be of use to you:

freq_max_amp Hz Frequency of the maximum energy of the audio event bin_max_energy Hz Frequency at the maximum of energy where the slope is the flattest

However, as you might have seen already, this function also stores all the amplitude / frequency information as lists allowing you to plot the power spectrum of any call that passed throught the detection filters. Perhaps the whole amplitude / frequency information of the call might allow you to calculate the values you are looking for...

Best, Francois Fabianek

Le dim. 22 août 2021 à 16:32, Dr Tom August @.***> a écrit :

While these two functions usually isolate the same calls, the metrics returned are quite different. This can be seen using the example data shipped with the package. I'm interested in the peak frequency/frequency centroid values, and while from the plots it is clear both functions have picked out the calls the metrics produced are very different. Do you know what is going on here/

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/WavX/bioacoustics/issues/15, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC5DBQVIXEO2T3R2TIVYOSDT6FNEFANCNFSM5CTJPPMA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email .