Multichannel Nonnegative Tensor Factorization With Structured Constraints For User-Guided Audio Source Separation

	Recording #1 (23 to 43 seconds snip)	Recording #2 (60 to 80 seconds snip)	Recording #3 (44 to 64 seconds snip)
Tracks separated by proposed method	Trumpet Bass Drums	Voice Sax Bass Piano Brushes	Voice Violin F.Horn Bass
Tracks separated by DUET [1]	Trumpet Bass+Drums	Voice+Sax Bass Piano+Brushes	-
Time codes (of full-length recordings)
Scatter plots (of full-length recordings)
Upmixing to 5.1 using the sources separated by the proposed method: C: center L/R: left, right Lf: low frequency Ls/Rs: L/R surround	C L R Lf Ls Rs	C L R Lf Ls Rs	C L R Lf Ls Rs
Videos from the project SARAH