Android 10/11 with Azure Speech - Workaround to record audio (Voice call) & pass chunks or stream audio (Voice call) to Azure Speech service

Jyotsna Ravi 51 Microsoft Employee

Android 10/11 requires specific permission for recording audio. Require help in recording the voice call or streaming the voice call in real time to Azure speech SDK. This was successfully implemented with Android 8.

Ashish 1 Reputation point

2022-03-25T10:11:16.297+00:00

Hi, I'm a developer from the above mentioned project, our objective is to transcribe voice calls in android. The problem in accessing audio during a call is due to android's tighter restrictions introduced for ensuring user privacy on android 10 and above. When we attempt to capture audio, the incoming voice just registers silence or in some cases - extremely damped audio. We have used Azure Speech SDK using the AudioRecord API for converting speech to text from the incoming call audio and it works well on android 8/9 but not on android 10/11. We cannot use VOICE_CALL audio source as android has restricted it's access to system apps only so we had to use VOICE_RECOGNITION which uses the device microphone to capture audio. The speaker's voice is getting processed and transcribed perfectly but it fails to capture the other person's voice. I have created a sample repository for explaining this issue with the respective stack traces.
GiftA-MSFT 11,161 Reputation points

2022-03-28T17:42:23.393+00:00

Thanks for providing more details. I'm making some inquiries, will provide updates as soon as possible.
GiftA-MSFT 11,161 Reputation points

2022-03-29T21:19:27.737+00:00

Started an email thread with the product team, let's continue offline. Thanks.

Accepted answer

GiftA-MSFT 11,161 Reputation points

2022-03-09T22:34:25.413+00:00

Hi, thanks for your feedback. You are correct, my previous response was about recognizing from microphone. The following document shows how to recognize from file. Instead of calling fromDefaultMicrophoneInput(), you call fromWavFileInput() and pass the file path. For your scenario, you'd want to consider using continuous recognition. I'm not aware of any limitations when using android 10/11 with speech sdk. If your question is specific to android 10/11 and recording audio, then maybe consider using 'dotnet-android' forum tag on this platform or post a question on android community forum for the community experts to review.
Please sign in to rate this answer.
Jyotsna Ravi 51 Reputation points Microsoft Employee

2022-03-23T03:54:56.757+00:00

Thanks for the reply. The customer wants to use Android 10/11 with speech SDK.
Can we get some help from Android experts from MS? There are some permissions required for recording or streaming voice calls on Android 10. Any work around scenarios which our engineers have come across would be a good pointer.

GiftA-MSFT 11,161 Reputation points

2022-03-23T20:05:15.22+00:00

Thanks for your feedback. I'm making some inquiries with engineers internally. It would be helpful if you can clarify what permission issues you're facing so I can pass that information along.
Sign in to comment

Share via

Android 10/11 with Azure Speech - Workaround to record audio (Voice call) & pass chunks or stream audio (Voice call) to Azure Speech service

0 additional answers