Skip to main content
EURASIP Journal on Advances in Signal Processing

Table 1 Instrumental measures for baseline and proposed approaches averaged over seen noise types of the development set data. Note that the SNR conditions of −5 and 15 dB are unseen during training, whereas the remaining SNRs have been seen. Best two approaches are in boldface

From: Speech enhancement by LSTM-based noise suppression followed by CNN-based speech restoration

SNR

Method

PESQ

STOI

SNRI

-5

Noisy

1.35

0.53

0.00

 

MMSE-LSA

1.39

0.50

3.51

 

SG-jMAP

1.38

0.49

4.13

 

LSTM-IRM

1.61

0.65

11.76

 

LSTM-MSA

1.65

0.66

12.61

 

LSTM-cMSA

1.60

0.64

15.83

 

CED-cSA-du

1.52

0.65

11.75

 

LSTM-cMSA + DNN-cSA

1.61

0.68

17.07

 

LSTM-cMSA + CED-cSA-du

1.63

0.69

17.12

 

LSTM-cMSA + CED-cSA-tr

1.63

0.69

17.27

0

Noisy

1.52

0.65

0.00

 

MMSE-LSA

1.64

0.63

4.37

 

SG-jMAP

1.63

0.63

5.19

 

LSTM-IRM

1.99

0.79

17.14

 

LSTM-MSA

2.08

0.80

19.21

 

LSTM-cMSA

2.03

0.80

24.59

 

CED-cSA-du

1.92

0.81

19.15

 

LSTM-cMSA + DNN-cSA

2.06

0.83

26.55

 

LSTM-cMSA + CED-cSA-du

2.12

0.85

26.10

 

LSTM-cMSA + CED-cSA-tr

2.12

0.84

26.50

5

Noisy

1.77

0.76

0.00

 

MMSE-LSA

1.97

0.75

5.10

 

SG-jMAP

1.98

0.75

6.14

 

LSTM-IRM

2.42

0.87

18.39

 

LSTM-MSA

2.53

0.88

20.20

 

LSTM-cMSA

2.55

0.89

26.13

 

CED-cSA-du

2.43

0.90

21.98

 

LSTM-cMSA + DNN-cSA

2.61

0.91

28.60

 

LSTM-cMSA + CED-cSA-du

2.70

0.92

27.79

 

LSTM-cMSA + CED-cSA-tr

2.70

0.92

28.47

10

Noisy

2.12

0.86

0.00

 

MMSE-LSA

2.36

0.84

5.54

 

SG-jMAP

2.41

0.85

6.88

 

LSTM-IRM

2.84

0.92

17.92

 

LSTM-MSA

2.93

0.93

19.22

 

LSTM-cMSA

3.01

0.93

25.51

 

CED-cSA-du

2.97

0.95

24.09

 

LSTM-cMSA + DNN-cSA

3.08

0.95

28.29

 

LSTM-cMSA + CED-cSA-du

3.17

0.96

27.17

 

LSTM-cMSA + CED-cSA-tr

3.19

0.96

28.16

15

Noisy

2.54

0.92

0.00

 

MMSE-LSA

2.76

0.91

5.53

 

SG-jMAP

2.87

0.92

7.22

 

LSTM-IRM

3.24

0.96

16.64

 

LSTM-MSA

3.32

0.96

18.20

 

LSTM-cMSA

3.39

0.96

23.47

 

CED-cSA-du

3.42

0.97

25.58

 

LSTM-cMSA + DNN-cSA

3.43

0.97

26.39

 

LSTM-cMSA + CED-cSA-du

3.51

0.97

25.03

 

LSTM-cMSA + CED-cSA-tr

3.54

0.97

26.31

Mean

Noisy

1.86

0.75

0.00

 

MMSE-LSA

2.02

0.73

4.81

 

SG-jMAP

2.05

0.73

5.91

 

LSTM-IRM

2.42

0.84

16.37

 

LSTM-MSA

2.50

0.85

17.89

 

LSTM-cMSA

2.52

0.85

23.11

 

CED-cSA-du

2.45

0.85

20.51

 

LSTM-cMSA + DNN-cSA

2.56

0.87

25.38

 

LSTM-cMSA + CED-cSA-du

2.62

0.88

24.64

 

LSTM-cMSA + CED-cSA-tr

2.63

0.88

25.34

Back to article page