Skip to main content

Table 2 The average PESQ scores with and without postfiltering for the typical DNN-based speech enhancement methods

From: Low-complexity artificial noise suppression methods for deep learning-based speech enhancement algorithms

SNR/dB − 5 0 5 10 Aver. SNR/dB − 5 0 5 10 Aver.
Noisy 1.32 1.66 2.00 2.40 1.85 Noisy 1.32 1.66 2.00 2.40 1.85
CRN 1.94 2.55 2.78 3.10 2.59 DCN 1.89 2.46 2.75 3.06 2.54
SPP-MMSE 2.00 2.59 2.83 3.19 2.65 SPP-MMSE 1.90 2.48 2.79 3.12 2.57
SPP-proposed-1 2.03 2.70 2.94 3.28 2.74 SPP-proposed-1 1.99 2.58 2.93 3.23 2.68
SPP-proposed-2 2.03 2.68 2.92 3.27 2.73 SPP-proposed-2 2.01 2.57 2.90 3.22 2.68
SPP-proposed-3 2.03 2.69 2.90 3.23 2.71 SPP-proposed-3 1.96 2.57 2.86 3.19 2.65
GRN 1.96 2.57 2.75 2.94 2.56 DARCN 1.97 2.65 2.86 3.15 2.66
SPP-MMSE 1.95 2.59 2.77 3.00 2.58 SPP-MMSE 1.96 2.70 2.90 3.22 2.70
SPP-proposed-1 2.03 2.71 2.89 3.12 2.69 SPP-proposed-1 2.00 2.80 3.00 3.33 2.78
SPP-proposed-2 2.03 2.71 2.86 3.11 2.68 SPP-proposed-2 1.96 2.74 2.95 3.27 2.73
SPP-proposed-3 2.01 2.70 2.85 3.06 2.66 SPP-proposed-3 1.99 2.80 2.96 3.31 2.77