Splitting FreeSWITCH audio recording into streams

[GyR4uK]

I would like to know about the experience of splitting audio streams into files when recording FreeSWITCH. If someone faced such a task, tell me how they implemented it.

This separation is required for further implementation of AI to recognize audio in text and analyze it.

[sd...@distancelearning.cloud]

Freeswitch has media bugs, so you can get the in/out audio of any user, and the mixed output of mod_conference which is sent to the html5 client.

There are some modules that exist in open source freewitch for speech to text, and they have commercial features in signalwire PaaS.

 

There are also some companies I work with that have 3d spatial stereo, and advanced noise cancellation plugins that couple with freeswitch and have been implemented commercially.   Some open source projects are attempting to do the same thing, but as you probably know, the neural models used are held closely.

 

Regards.

Stephen

--
You received this message because you are subscribed to the Google Groups "BigBlueButton-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bigbluebutton-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bigbluebutton-dev/32244fc1-559f-4135-9681-6b2ebffc1bd0n%40googlegroups.com.

[GyR4uK]

I completely agree with you. For the time being, it is planned to use a commercial AI model for recognition. The biggest difficulty so far has caused me to correctly save the audio tracks of users from the meeting.

Namely:

1) You need to put all the records in a specific folder with recordId or meetingId.

2) Name the files with the username or userID

среда, 17 марта 2021 г. в 01:27:27 UTC+3, DistanceLearning.cloud:

[Robert]

When searching for commercial solutions there is a unimrcp module in Freeswitch.
With the media bugs you can name the files automatically as "username.wav" for example or use the Freeswitch uuid. I don't know if Freeswitch support creating folders but with a bit of scripting it is possible.