AVAudioEngine in a VoIP app

Hello,


Our used case is real-time audio (voip) application. I wanted to use AVAudioEngine for to capture and render audio.


This is how I am setting up the AVAudioEngine nodes in my app -

AVAudioUnit -> MainMixer -> Speaker

Microphone -> InputNode -> MicMixer -> Tap (write audio to a file)


When I am writing the tapped buffers into a file, I noticed, the audio played on speaker gets captured by the microphone and gets written to file. I do not want the audio played on the speaker to be captured.


Is Acoustic Echo Cancellation (AEC) is supported by AVAudioEngine? Do I need to make any specific configuration or design changes to achieve the AEC? Is there a recommendation on how to achieve the mentioned used case?


Any help for this is greatly appreciated.

Thank you,

Piyush

Up vote post of ptank Down vote post of ptank
5.2k views
Posted by
ptank
Reply
Add a Comment

Replies

Hello,


I was able to achieve bi-directional audio with echo cancellation using VoiceProcessingIO Audio Unit and AVAudioEngine's manual rendering mode.


Example code: https://github.com/twilio/video-quickstart-swift/blob/master/AudioDeviceExample/AudioDevices/ExampleAVAudioEngineDevice.m


The example uses CoreAudio's VoiceProcessingIO audio unit to playback and records audio at up to 48KHz with built-in echo cancellation. The example uses AVAudioEngine in manual rendering mode for mixing the Remote Participant's audio and audio from a file. CoreAudio receives the mixed audio samples from the AVAudioUnit's output node.


Thank you,

Piyush

Posted by
ptank
Up vote reply of ptank Down vote reply of ptank
Add a Comment

Amazing !


I searched for using Acoustic Echo Cancellation (AEC) with AudioEngine for a very long time but I didn't find anything.

I tried using https://developer.apple.com/documentation/avfoundation/avaudiounit but without success.


Thank you for your sample !

Posted by
gbitaudeau
Up vote reply of gbitaudeau Down vote reply of gbitaudeau
Add a Comment