(I started an earlier thread in the Audio SDK forum, but this issue is unrelated to spatial audio so I'm posting in general support.)
We're making SoundSelf, an accessible voice-driven experience for Vive. The player interacts with the game by making sustained tonal sounds with their voice. However, Vive's noise suppression algorithm identifies any sustained sound (>.75s) as background noise, and cuts it (-40-60dB) from the signal. In our case, this means Vive is deleting player input — making our experience unplayable.
Like most voice-recognition technologies, our technology works best with unprocessed audio. In order to support voice-recognition apps better, Windows 10 includes a "raw mode" in its WASAPI audio API. Oculus implements raw mode in its mic audio driver, allowing us to easily circumvent its noise suppression. Vive, however, does not — and I have been unable to find any API call or user preference to do so.
This is truly a show-stopping issue for us and we may not be able to ship on Vive if it can't be addressed. It's like playing a game where all the buttons and joysticks automatically reset to zero every split-second.