Microsoft has trained its algorithm to identify more than 150 annoying background noises and stop them from interrupting your calls.
Microsoft has introduced some clever technology into Teams that could make those video calls a little less chaotic.
users now have the option to switch on a new noise suppression feature, which uses machine learning to detect unwanted background noise (barking dogs, overexcited children or relatives who have gotten a little too merry, for example) and filter it out, keeping conversations clear.
The technology works by analyzing each caller’s audio feed and identifying unnecessary noise based on a dataset comprising approximately 760 hours of “clean” speech and 180 hours of noise data. The algorithm is supposedly able to identify 150 different noises, including clacking keyboards, flushing toilets and snoring, and suppress them from the audio feed.
The result is that, even if there are distracting noises on one caller’s end, the person on the other side of the screen won’t be able to hear them.
which is currently available on version 5.2.0 and higher of the Zoom desktop client for Windows. According to Microsoft, while traditional noise suppression algorithms can only suppress static noises (like the noise from a fan), its technology can also pick out “non-stationary noises” that drop in and out of the environment.
At the same time, the algorithm has been trained on a wide variety of voices, languages and emotions, to ensure the noise suppression feature doesn’t unintentionally mute voices it doesn’t recognize.
Microsoft explained: “For clean speech we ensured that we had a balance of female and male speech and we collected data from 10+ languages which also include tonal languages to ensure that our model will not change the meaning of a sentence by distorting the tone of the words.
“Another important aspect was to include emotions in our clean speech so that expressions like laughter or crying are not suppressed.”
The wide variety of environments that Teams users dial in from also has a big impact, Microsoft said – no house is the same, after all. To capture this diversity, Microsoft trained its model with data from more than 3,000 real-room environments and more than 115,000 “synthetically created” rooms.
The algorithm was then put through its paces with human listeners, who were instructed to score the quality of their audio on a scale of one to five stars. “With these human ratings we were able to develop a new perceptual metric which together with the subjective human ratings allowed us to make fast progress on improving the quality of our deep learning models,” said Microsoft.
The new noise suppression feature is currently only available in the Windows desktop version of Microsoft Teams, and can be enabled by navigating to the Noise suppression drop-down in the device settings and selecting ‘High’. You can see it in action on Microsoft’s blog post.
Microsoft is now working on bringing the feature to Mac and mobile platforms.
Last week, the company introduced
to Microsoft Teams.