Skip to main content

Table 8 PESQ values of baselines under unseen noise condition. Proposed model represented by bold and italic letters

From: Sub-convolutional U-Net with transformer attention network for end-to-end single-channel speech enhancement

Metric

PESQ

Noise

Train

Airport

Exhibition hall

SNR (dB)

− 5

0

5

Avg.

− 5

0

5

Avg.

− 5

0

5

Avg.

Noisy mixture

1.17

1.41

1.74

1.44

1.31

1.67

1.93

1.64

1.64

1.88

2.15

1.89

Bi-LSTM [31]

1.85

2.16

2.49

2.17

1.98

2.26

2.58

2.27

2.18

2.31

2.66

2.38

Bi-CRN [34]

1.94

2.21

2.57

2.24

2.06

2.34

2.64

2.35

2.27

2.44

2.77

2.49

SEGAN [40]

2.02

2.32

2.66

2.33

2.19

2.46

2.75

2.47

2.38

2.57

2.93

2.63

GRN [30]

2.08

2.44

2.71

2.41

2.26

2.55

2.84

2.55

2.47

2.63

3.06

2.72

DCN [38]

2.17

2.59

2.85

2.54

2.34

2.61

2.96

2.64

2.55

2.77

3.14

2.82

DCCRN [35]

2.29

2.65

2.94

2.63

2.42

2.69

3.04

2.72

2.61

2.86

3.20

2.89

TSTNN [41]

2.36

2.71

3.01

2.69

2.51

2.76

3.11

2.79

2.69

2.93

3.24

2.95

MASENet [46]

2.42

2..77

3.08

2.75

2.63

2.84

3.18

2.88

2.76

3.01

3.29

3.02

SADNUNet [47]

2.55

2.83

3.17

2.85

2.71

2.92

3.26

2.96

2.81

3.06

3.33

3.07

MCGN [42]

2.61

2.89

3.21

2.90

2.78

3.04

3.34

3.05

2.85

3.16

3.42

3.14

DBT-Net [51]

2.67

2.92

3.27

2.96

2.82

3.09

3.37

3.09

2.89

3.24

3.49

3.20

TANSCUNet

2.89

3.07

3.46

3.14

2.93

3.25

3.58

3.25

3.01

3.35

3.62

3.33