Skip to main content
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: IEEE/ACM Trans Audio Speech Lang Process. 2021 Mar 8;29:1270–1279. doi: 10.1109/taslp.2021.3064421

TABLE I:

Performance comparisons between different configurations of dense block, dilation, and attention in DCN. Boldface indicates the best score in a given condition.

Metric STOI PESQ SNR
Test noise Babble Cafeteria Babble Cafeteria Babble Cafeteria
Test SNR (dB) −5 0 5 Avg. −5 0 5 Avg. −5 0 5 Avg. −5 0 5 Avg. −5 0 5 Avg. −5 0 5 Avg.
Mix. m ↓ Dil. Att. 58.4 70.5 81.3 70.1 57.1 69.7 81.0 69.2 1.56 1.82 2.12 1.83 1.46 1.77 2.12 1.78 −5.0 0.0 5.0 0 −5.0 0.0 5.0 0.0
Causal 1 76.7 88.0 93.2 86.0 76.4 87.8 92.9 85.7 1.90 2.39 2.76 2.35 2.02 2.49 2.84 2.45 5.5 9.9 13.4 9.6 6.5 10.4 13.4 10.1
2 81.6 91.3 95.0 89.3 80.5 90.2 94.3 88.3 2.13 2.70 3.08 2.64 2.17 2.68 3.05 2.63 7.4 11.5 14.7 11.2 7.7 11.4 14.4 11.2
2 83.5 91.9 95.2 90.2 81.4 90.5 94.5 88.8 2.23 2.75 3.12 2.70 2.21 2.70 3.07 2.66 7.7 11.8 15.0 11.5 7.9 11.5 14.5 11.3
2 84.9 92.2 95.3 90.8 82.1 90.7 94.6 89.1 2.30 2.77 3.14 2.74 2.23 2.71 3.08 2.67 8.2 12.0 15.1 11.8 8.2 11.7 14.7 11.5
2 85.3 92.3 95.4 91.0 82.3 90.8 94.7 89.3 2.34 2.81 3.17 2.77 2.24 2.72 3.09 2.68 8.5 12.1 15.1 11.9 8.2 11.7 14.7 11.5
1 83.9 91.8 95.2 90.3 81.0 90.3 94.5 88.6 2.23 2.72 3.09 2.68 2.15 2.62 3.01 2.59 7.9 11.8 15.0 11.6 7.9 11.5 14.5 11.3
Non-causal 3 84.7 92.5 95.7 90.9 83.1 91.4 95.0 89.8 2.37 2.88 3.22 2.82 2.34 2.82 3.16 2.77 8.2 12.2 15.2 11.9 8.3 11.8 14.7 11.6
3 86.6 92.9 95.7 91.7 84.1 91.7 95.0 90.3 2.53 2.96 3.24 2.91 2.44 2.88 3.19 2.84 9.1 12.5 15.3 12.3 8.7 12.0 14.8 11.8
3 87.9 93.5 96.0 92.4 85.0 92.0 95.2 90.8 2.61 3.02 3.32 2.98 2.47 2.91 3.24 2.87 9.6 12.9 15.7 12.7 8.9 12.2 15.0 12.0
3 87.9 93.5 96.1 92.5 85.0 92.1 95.3 90.8 2.61 3.04 3.33 2.99 2.45 2.91 3.23 2.86 9.6 12.9 15.8 12.8 8.9 12.3 15.1 12.1
1 83.7 91.5 95.2 90.1 80.1 89.8 94.3 88.1 2.24 2.71 3.09 2.68 2.13 2.59 2.98 2.57 8.3 12.0 15.2 11.8 7.8 11.4 14.6 11.3