6
$\begingroup$

Background:

I've been asked to analyze a 2-channel (which I presume means stereo) .wav sound file to identify a single bird species with BirdNET (I'm not sure of the software version, but the GUI shows a 2022 copyright).

The process works fine, outputting to a Raven Pro selection table.

However, I notice in the output selection table that the column Channel is always populated with the value 1, as shown in this screenshot:

enter image description here

Out of curiosity, I wondered if I could instead output channel 2, but I could not find that option, either in the BirdNET interface (GUI or non-GUI), or in the documentation.

Question:

How to force BirdNET to output either channel 1 or 2, when using a 2-channel input sound file?

$\endgroup$
3
  • $\begingroup$ I searched the BirdNET-Analyzer Github repository for insight around channel selection, and didn't find anything. I usually use the R output type instead of Raven selection table, and channel is not even a column in the R output type. It's also not an input command line argument. This makes me suspect channel 1 is a default for BirdNET, and perhaps the channel column only exists for compatibility with Raven Pro. Your best bet may be to ask your question of the BirdNET development team directly by submitting an issue here: github.com/kahst/BirdNET-Analyzer/issues $\endgroup$
    – Cathleen B
    Jun 8, 2023 at 17:27
  • $\begingroup$ Thanks, Cathleen. Your comment makes sense, and if you'd like to turn it into an answer I will give you credit. At your suggestion, I've posted to GitHub at: github.com/kahst/BirdNET-Analyzer/issues/144 $\endgroup$
    – Stu Smith
    Jun 13, 2023 at 17:45
  • $\begingroup$ Hi Stu, if the BirdNET team's response answers your question, I think you can actually summarize that here as an answer to your own question, and accept it! (Moderators, please correct me if I'm mistaken.) $\endgroup$
    – Cathleen B
    Jun 13, 2023 at 22:18

1 Answer 1

3
$\begingroup$

Here is the response to my inquiry to the BirdNET development team at Github:

BirdNET uses the librosa library to load the audio files, its load function has the option to convert the signal to mono after reading. You can find the source here, but basically the audio signal is converted by averaging samples across channels.

So, to answer my own question, you cannot specify a specific channel for BirdNET analysis and output. Further, if the input sound file has multiple channels BirdNET will average them into one.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.