Yealink provides a full range of Microsoft Teams devices. MSpeech is an intelligent USB speakerphone certified by Microsoft Teams. It features 10 MEMS microphone arrays and 4-watt speakers to provide AI features, including voice recognition, transcription, translation, and more intelligent experiences for MicrosoftTeams Rooms.

This article mainly explains how to use MSpeech to identify in-room participants and provide live transcription.

Tenant Settings

Before the deployment, we recommend that you read this Microsoft doc to ensure that your tenant settings are ready.

https://docs.microsoft.com/en-us/microsoftteams/rooms/voice-recognition

Connection and Setup

Connection

Yealink MSpeech is connected to the MTouch II via a USB cable to get the power supply from MTouch II and transmit audio data with MTouch II. After connection, please choose UAC2_RENDER as speaker and UAC2_TEAMS as microphone in the device options settings.

Upload Your Voice Profile on Teams PC Client

When uploading the voice profile, make sure the app language is English (United States) so that you can find the Recognition in Teams app settings. You will need to read the on-screen text to finish this step.

Identify In-Room Participants during Meetings

You can start the transcription in MTR or Teams PC Client. If the MTR with MSpeech is in the meeting, you can see a prompt showing “XXX started identifying people”.

Suppose you have uploaded your voice profile and been invited to a meeting. In that case, your speech will be automatically recognized and transcribed when you start talking with an intelligent speakerphone in the meeting. Your transcript appears on the meeting video in real time. The transcripts include your name and a timestamp, as shown in the picture below:

Suppose you are invited to a meeting but have not uploaded your voice profile. In that case, your speech will be automatically recognized and transcribed when you start talking with an intelligent speakerphone in the meeting.

However, your transcript will be recognized as “SpeakerX”.

Select Identify speaker and assign this speech to the desired participant from the search list (every invitee will appear on the list). The following transcripts will include the participant’s name and a timestamp.

Notes: Only invitees from the same tenant can be selected in the identity speaker drop-down list.

Things Need to Know in the Actual Demo and Setup

1.         Live transcription can be used in all MTR systems, and the MSpeech brings the participant identifying capability into the MTR system. Microsoft defines this feature as a voice attribution transcription service. Currently, this service is only available for US Tenant users and will expand to non-US-based customers in April 2022. This expansion also supports voice attribution transcription services for 16 different languages, including German, Simplified Chinese-Mandarin, Japanese, Brazilian Portuguese, Spanish, Korean, French, and Italian, with more to come.

2.         You should use the Meeting Room Standard or Meeting Room Premium account instead of the E3/E5 account.

3.         The meeting should be invitation-type. There should be less than 20 people in the email invite, and the in-room participants should be less than ten people.

4.         Each in-room participant must be invited individually, on the original or through a forwarded invitation.

5.         Only the people in the same tenant with the person who starts the transcription are identified by Intelligent Speakers.