Skip to main content

Table 2 Results when the spatial filter output is multiplied by the estimated DSPP

From: DOA-informed source extraction in the presence of competing talkers and background noise

 

DSB-inst

DSB-cPSD

\(\mathcal {D}_{\text {dm}}\)-inst

\(\mathcal {D}_{\text {dm}}\)-cPSD

NR

14.3

14.2

19.3

18.9

IR

15.7

15.6

24.9

24.3

ν sd

0.36

0.36

0.33

0.34

Δ PESQ

0.52

0.47

0.88

0.79

Δ STOI

0.10

0.08

0.13

0.11

NR

9.1

8.2

15.7

15.4

IR

12.3

11.5

25.0

23.1

ν sd

0.12

0.13

0.15

0.15

Δ PESQ

0.67

0.60

1.07

1.00

Δ STOI

0.08

0.07

0.11

0.10

NR

12.2

11.4

16.7

16.0

IR

14.0

13.2

22.5

21.6

ν sd

0.27

0.26

0.24

0.23

Δ PESQ

0.54

0.48

0.87

0.84

Δ STOI

0.04

0.02

0.05

0.04

  1. Source1 (top), Source2 (middle), and Source3 (bottom). The segmental DSIR at the reference microphone of each source is 6.8, 5.7, and 8.0 dB, "-inst" indicates the DOA estimator with instantaneous phase differences, while “-cPSD” the one with cross-PSD phase differences. The best result is indicated in bold