Skip to main content

Table 3 Detailed results obtained by DS+DNN8 on the six test conditions

From: Speech dereverberation for enhancement and recognition using dynamic features constrained deep neural networks and feature adaptation

Cases

Simulated rooms

Real

 

CD (dB)

SRMR

LLR

SNR (dB)

SRMR

far1

1.64/1.52

4.85

0.12/0.09

14.33/14.52

4.81

far1

2.76/2.49

4.83

0.31/0.24

10.18/11.36

-

far1

2.44/2.21

4.16

0.28/0.22

10.30/11.13

-

near1

1.44/1.33

4.72

0.09/0.07

15.15/15.34

4.43

near1

2.00/1.82

4.60

0.17/0.12

13.04/13.64

-

near1

1.97/1.80

4.49

0.19/0.14

12.82/13.30

-

avg.

2.04/1.86

4.61

0.193/0.147

12.64/13.22

4.62

  1. For CD, LLR, and SNR, the numbers before and after “/” represents the mean and median of the scores, respectively