Skip to main content

Advertisement

Table 4 Reconstruction SDR of audio source separation in decibel. M/V represents the energy ratio between music and voice

From: Discriminative frequency filter banks learning with neural networks

M/V 0.1 1 10
TriFB 4.47 8.30 12.01
GaussFB 4.85 8.39 12.22
TriFB-DN 5.19 8.51 12.92
GaussFB-DN 5.13 8.45 13.01
BandPosFB-DN 5.33 8.39 12.84
PosFB-DN 5.70 9.14 16.99