hope to see the support of music protection in the process of AI noise suppression

[zuowanbushiwo]

In some scenarios, music need to be preserved during video conferencing. Since the characteristics of some music elements and noise are very close, if the deep learning noise reduction model is directly applied, the music will be suppressed very stuck. If music protection is added to model training, the effect of noise suppression in other corpora will be affected.
So I think it might be a good choice to add a music event detection module, it can identify whether the sound is music or noise. After the music is recognized, the noise reduction can be turned off without affecting the noise reduction in normal scenes. Thanks！

[robin1001]

Got it!