Speech-to-text detect speakers while using microphone

Question

Speech-to-text detect speakers while using microphone

Hey there, we are facing an issue when using speech to text in realtime steam. The microphone always pick both phone-speakers voice and voice from the person who is speaking, then convert the mixed voice to text. We want to add an interrupt mechanism - when it detects user start to speak, the phone-speaker could automatically switches to silent mode. But because of the mixed voices, it can not be done.

Is there any VAD mode or speech activity detection services in Azure we can use, like it can detect how many people are speaking, or other ways to resolve this?

Ramr-msft 17,826 Reputation points

2023-07-27T17:50:02.6733333+00:00

@海南好视花声文化传媒有限公司 Thanks for the question, Yes, Azure Speech Services provides a Speech-to-Text API that includes a Voice Activity Detection (VAD) feature. The VAD feature can be used to detect when someone is speaking and filter out background noise or other voices.

Here are the high-level steps to follow:

Use the Azure Speech Services Speech-to-Text API to capture audio from the microphone in real-time.

Enable the VAD feature in the Speech-to-Text API to detect when someone is speaking.

Use the VAD output to trigger the interrupt mechanism to switch the phone-speaker to silent mode.
海南好视花声文化传媒有限公司 0 Reputation points

2023-08-02T02:35:52.46+00:00

Thanks for the feedback @Ramr-msft ! We tried to locate the feature but couldnt find where to enable the VAD feature in the documentation, could you share the guildline url about it?
海南好视花声文化传媒有限公司 0 Reputation points

2023-08-02T02:38:10.45+00:00

Thanks for the quick feedback @Ramr-msft ! We tried to locate the feature but couldnt find in the documentation about where/how to enable VAD, could you kindly share the guideline url?

Your answer

Ramr-msft 17,826 Reputation points

2023-07-27T17:50:02.6733333+00:00

@海南好视花声文化传媒有限公司 Thanks for the question, Yes, Azure Speech Services provides a Speech-to-Text API that includes a Voice Activity Detection (VAD) feature. The VAD feature can be used to detect when someone is speaking and filter out background noise or other voices.

Here are the high-level steps to follow:

Use the Azure Speech Services Speech-to-Text API to capture audio from the microphone in real-time.

Enable the VAD feature in the Speech-to-Text API to detect when someone is speaking.

Use the VAD output to trigger the interrupt mechanism to switch the phone-speaker to silent mode.
海南好视花声文化传媒有限公司 0 Reputation points

2023-08-02T02:35:52.46+00:00

Thanks for the feedback @Ramr-msft ! We tried to locate the feature but couldnt find where to enable the VAD feature in the documentation, could you share the guildline url about it?
海南好视花声文化传媒有限公司 0 Reputation points

2023-08-02T02:38:10.45+00:00

Thanks for the quick feedback @Ramr-msft ! We tried to locate the feature but couldnt find in the documentation about where/how to enable VAD, could you kindly share the guideline url?

Share via

Speech-to-text detect speakers while using microphone

Your answer