hi all!
i have a short question. the "original" definition of SNR from 2006 (by E. Vincent) - https://hal.inria.fr/inria-00544230/document - is about ||s_target + e_inferf|| / ||e_noise||.
but your book https://source-separation.github.io/tutorial/basics/evaluation.html#sdr-sir-and-sar says it's || s_target || / || s_target - s_hat ||, which is equal to || e_interf + e_noise + e_artif||, which would make SNR == SDR.
and i'm confused even more when i saw nussl follows the same definition of your book.
am i missing something?
hi all!
i have a short question. the "original" definition of SNR from 2006 (by E. Vincent) - https://hal.inria.fr/inria-00544230/document - is about ||s_target + e_inferf|| / ||e_noise||.
but your book https://source-separation.github.io/tutorial/basics/evaluation.html#sdr-sir-and-sar says it's || s_target || / || s_target - s_hat ||, which is equal to || e_interf + e_noise + e_artif||, which would make SNR == SDR.
and i'm confused even more when i saw
nusslfollows the same definition of your book.am i missing something?