2 Replies
      Latest reply on Apr 16, 2019 8:29 PM by jackyyigao
      lhall Level 1 Level 1 (0 points)

        Hi all,


        We have a library that uses WebRTC (webrtc.org) to offer voice and video communication from iOS applications, amongst other platforms. One of our customers is using this library in an audio translation use-case. The iOS app is in a VoIP call with a remote browser (Chrome on Windows). The person on the Chrome end is providing live translation for the people on the iPad end.


        Our customer is reporting that when both parties are speaking at the same time, there is a "muting" effect from the iPad - in other words, the interpreter on Chrome momentarily loses the audio from the iPad.


        The iPad is attached to a loudspeaker so that more than one person in the room can take part in the conversation.


        We are trying to find a way to eliminate the muting effect. Our assumption so far is that the muting is caused by the echo cancellation in the Voice Processing IO audio unit that is used in WebRTC. By patching WebRTC to replace the VPIO audio unit with the Remote IO audio unit, we can eliminate the muting, but we introduce echo on the call. We also find that the mic gain is much lower with the Remote IO audio unit.


        Is there any way to configure the audio unit to eliminate the muting whilst also maintaining effective echo cancellation?




        • Re: VoIP muting in "double-talk" scenario
          arshiacont Level 1 Level 1 (0 points)

          Hi Leigh,


          Did you find any answer to your question?


          I am also experiencing the Mute effect only on iPad whereas echo cancellation seems to work properly (without muting) on iPhone.





            • Re: VoIP muting in "double-talk" scenario
              jackyyigao Level 1 Level 1 (0 points)

              I met this issue also. this seems not an easy question, echo cancellation often works together with some "non-linear processing" module to achieve better EC performance. that means the audio level is adjusted non linearly and sometimes greatly suppressed if the algorithm thinks there should be only one person is speaking at one end alghought acutally there are voice from both ends.