search

introlab / odas

ODAS: Open embeddeD Audition System
MIT License
780 stars 248 forks source link

classify #35

Closed taospartan closed 6 years ago

taospartan commented 6 years ago

What does the classify section of the config do?

Many thanks

FrancoisGrondin commented 6 years ago

The classify section aims to tell if the separated sound source is speech or another daily event sound (door slam, phone ringing, etc.) Unfortunately, the method only works when a single source is active, and still needs to be improved. It is included in the ODAS library, but not supported yet. We may add some neural networks to improve robustness. See this paper that explains the current method used: https://pdfs.semanticscholar.org/fb68/b776f00740b715d4bd2744ee86258498ad13.pdf

taospartan commented 6 years ago

Thanks for the information, an interesting read!