Table 8 Denoising performance comparison of CTFUNet with other models
From: Channel and temporal-frequency attention UNet for monaural speech enhancement
Model | #Para. | Year | With reverb | Without reverb | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Â | Â | Â | WB-PESQ | NB-PESQ | STOI(%) | SI-SDR | WB-PESQ | NB-PESQ | STOI(%) | SI-SDR |
Noisy | - | - | 1.822 | 2.753 | 86.62 | 9.033 | 1.582 | 2.454 | 91.52 | 9.07 |
DCCRN [51] | 3.7M | 2020 | - | 3.077 | - | - | - | 3.266 | - | - |
DCCRN+ [52] | 4.7M | 2021 | - | 3.30 | - | - | - | 3.33 | - | - |
Conv-TasNet [8] | 5.1M | 2019 | 2.75 | - | - | - | 2.73 | - | - | - |
PoCoNet [53] | 50M | 2020 | 2.832 | - | - | - | 2.748 | - | - | - |
CTS-Net [54] | 4.4M | 2021 | 3.02 | 3.47 | 92.7 | 15.58 | 2.94 | 3.42 | 96.66 | 17.99 |
FullSubNet [14] | 5.6M | 2021 | 3.057 | 3.584 | 92.11 | 16.04 | 2.882 | 3.428 | 96.32 | 17.30 |
GaGNet [55] | 5.9M | 2022 | 3.18 | 3.57 | 93.22 | 16.57 | 3.17 | 3.56 | 97.13 | 18.91 |
FullSubNet+ [15] | 8.7M | 2022 | 3.177 | 3.648 | 93.64 | 16.44 | 3.002 | 3.503 | 96.67 | 18.00 |
FS-CANet [56] | 4.2M | 2022 | 3.218 | 3.665 | 93.93 | 16.82 | 3.017 | 3.513 | 96.74 | 18.08 |
CTFUNet | 6.1M | 2023 | 3.367 | 3.741 | 94.39 | 17.16 | 3.176 | 3.639 | 97.17 | 18.66 |