You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I tried to use the test script for restoring my audio, but I obtained worse performance.
I suspect the model only supports target sound from close field.
The text was updated successfully, but these errors were encountered:
Yes, some far-field voices may be treated as noise by the model. And during training, the distance between source and microphone is sampled in this Gaussian distribution: N(μ, σ2), μ = 2, σ = 4, which is usually not that far. So the model may become worse for the far-field voice.
In my case, it works for test speech from 1 meter distance. But it seems worse performance for test speech from 2 meter distance. More distortion is found in processed target speech while less reverberation and noise are gained.
I tried to use the test script for restoring my audio, but I obtained worse performance.
I suspect the model only supports target sound from close field.
The text was updated successfully, but these errors were encountered: