(Figure from:Microsoft Research)
Virtual microphone arrays can be used with existing devices (such as those with regular microphones)Mobile phoneOr laptops, etc.), dynamically combined into a larger array.
Specifically, the system is implemented based on techniques such as continuous audio stream alignment, blind beamforming, speech recognition, Diarization, and system fusion.
With seven sets of audio stream inputs,MicrosoftThe Institute has achieved a typos rate of 22.3%. In the non-overlapping segments of the close-range microphone, the erroneous rate can be controlled within 3%.
The number of users supported by the system depends on the number of microphones in the area. Microsoft pointed out that the technology makes it easy for customers to transcribe session content anytime, anywhere, regardless of whether they have a dedicated microphone array.