Microsoft Teams Use Machine Learning To Add New Audio Features And Improve Meetings

This Article is written as a summary by Marktechpost Staff based on the Microsoft article. All Credit For This Research Goes To Researchers on This Project. 

Please Don't Forget To Join Our ML Subreddit

Online conversations and meetings can be less productive when there are distracting echo effects, poor room acoustics, and unstable visuals, to name a few frequent problems. Microsoft introduced new upgrades in Microsoft Teams to handle audio and video difficulties in both user-friendly and scalable ways across contexts, thanks to AI and machine learning, which have become core to our approach for continuous development.

Microsoft hopes to provide users with a better experience by enhancing Team audio and video quality. Recently, the tech giant revealed many improvements for its video calling platform that do the same. Meetings and conference calls will seem more natural because of Microsoft Teams’ use of artificial intelligence (AI). Echo cancellation, “de-reverberation,” background noise reduction, and real-time optimization are some of the solutions that leverage AI to improve user experiences that are now accessible.

Improvements to voice quality

Cancellation of echo

It’s typical for sound to loop between input and output devices during calls and meetings when a participant’s microphone is too close to the speaker, creating an unwelcome echo effect. Microsoft Teams can now detect the difference between a speaker’s sound and the user’s voice and remove the echo without muting the conversation or preventing many speakers from speaking simultaneously.

Poor room acoustics are compensated for by “de-reverberation.”

The acoustics of a room can cause sound to bounce or rebound in specific settings, giving the user’s voice the impression that they are speaking in a cavern. For the first time, Microsoft Teams converts collected audio signal to sound like users are talking into a close-range microphone using a machine learning model.

For more natural talks, interruptions.

The capacity to pause a discussion to ask a question or confirm something is a natural part of the discourse. Full-duplex (two-way) audio transmission is used to do this, enabling users to converse and hear one another simultaneously. It is challenging to eliminate echo while preserving full-duplex audio when not using a headset, particularly when utilizing equipment where the speaker and microphone are placed near one another. To maintain preferred voices while suppressing undesired audio signals, Microsoft Teams utilizes a model that has been “trained” using 30,000 hours of speech samples. This results in more fluid conversation.

Suppression of background noise

Every one of us may recall when a conference was abruptly ended by the sounds of a barking dog, a vehicle alarm, or a slammed door. We revealed the availability of AI-based noise reduction in Microsoft Teams as an optional feature for Windows customers more than two years ago. To further refine our model, we have since kept up an iterative development, testing, and assessment cycle. We have made machine learning-based noise suppression available by default for Teams clients using Windows (including Microsoft Teams Rooms), Mac, and iOS after observing notable improvements across key user metrics. The Teams Android and web applications will get a future update of this capability.

Enhanced video quality

Microsoft has also unveiled improved AI-based video and screen sharing quality enhancement for Teams. Companies increasingly use AI to help you appear and show your best, from low light modifications to optimizations based on the sort of information being shared.

The material you’re sharing is taken into account by real-time screen optimization.

The capacity of an audience to read on-screen text or view a shared video may frequently affect how effective a presentation is. However, different shared material types call for different strategies to guarantee the best possible video quality, especially with bandwidth restrictions. To improve the readability of texts or the smoothness of video playing, teams increasingly utilize machine learning to identify and modify the features of the information shown in real-time.

Even with limited bandwidth, your video will still look excellent, thanks to AI-based optimization.

Unexpected network capacity problems might result in a choppy video, which can rapidly cause your presentation’s attention to change. Teams’ AI-driven optimizations adapt playback in low bandwidth situations, allowing presenters to use video and screen sharing without concern.

Brightness and focus filters that make you appear as bright as possible

New AI-powered filters in Teams allow you the choice to modify brightness and add a soft focus for your meetings with a simple toggle in your device settings to better compensate for low-light circumstances, even though you can’t always control the lighting for your sessions.

Conclusion:

Microsoft introduced new upgrades in Microsoft Teams to handle audio and video difficulties using AI and machine learning. Microsoft Teams can now detect the difference between a speaker’s sound and the user’s voice and remove the echo without muting the conversation or preventing many speakers from speaking simultaneously.

Microsoft has also unveiled improved AI-based video and screen sharing quality enhancement for Teams. Teams’ AI-driven optimizations adapt playback in low bandwidth situations, allowing presenters to use video and screen sharing without concern.