LRS3-For-Speech-Separation
LRS3-For-Speech-Separation copied to clipboard
how to get mix_2_spk_cv.txt
how to get mix_2_spk_cv.txt ,how to get mix_2_spk_tr.txt and how to get mix_2_spk_tt.txt? How does the weighting coefficient of the two audio mixes come from? Why are the coefficients of the two audio frequencies just opposite?
I don't know if you have ever understood the task of pure speech separation. Depending on the task setting it is usually a mix of SNR between -5 and 5. The coefficients I have in the .txt file are randomly selected between -5db and 5db as SNR weights to mix the speech signals of different speakers. And I made sure that the same speaker does not exist in the same mixture, and also A+B and B+A do not appear at the same time.