Table 4 Comparison between methods based on deep learning techniques and our proposed variants in terms of SDR, SIR, and SAR with respect to the number of sources in the mixture

Exp-Wave-U-Net

2

− 0.42

1.75

10.98

3

− 3.85

− 2.74

11.97

4

− 5.90

− 5.33

12.87

scoreSynthFix

2

4.65

8.82

9.22

3

4.23

8.30

7.86

4

2.23

4.57

5.54

CExp-Wave-U-Net

2

− 0.16

4.62

7.48

3

− 0.68

2.88

5.91

4

− 2.56

0.44

6.35

CNN

2

5.00

7.66

8.74

3

3.03

5.47

6.03

4

1.83

1.37

5.52

scoreSynthFree-DTWoff

2

8.21

16.86

10.15

3

6.46

14.16

8.03

4

4.65

11.79

5.77

scoreSynthFix-DTWon

2

7.60

16.06

9.69

3

6.20

13.39

7.84

4

4.54

11.01

5.70