Skip to main content

Table 1 Performance comparison

From: Deep encoder/decoder dual-path neural network for speech separation in noisy reverberation environments

Separator network

Model size

(M)

MACs

(G)

SISNRi

(dB)

SDRi

(dB)

STOI

Conv-TasNet

5.1

20.8

7.4

7.9

0.71

DPRNN-TasNet

2.6

85.0

10.3

10.9

0.81

DPTNet

2.6

208.4

12.2

12.7

0.84

Proposed method

3.8

189.3

12.4

12.8

0.87

 Pro

w/o deep encoder/decoder

2.6

-

11.3

11.8

0.85

 Pro

w/o \({\mathcal{L}}_{\mathrm{STOI}}\)

3.8

-

12.3

12.7

0.84

 Pro

w/o SOSISNR

w SISNR

3.8

-

11.7

12.0

-