Speech Enhancement on VCTK Vibration sensor 12-bit, 4-16 kHz upsampling (test)
0.84LSD (Log-Spectral Distance)SUBARU
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SUBARUDevice=Desktop, Inference time (ms)=1.742025.06 | 0.84 | 4.35 | 4.19 | 17.94 | 3 | 90 | |
| SUBARUDevice=Pixel7, Inference time (ms)=70.412025.06 | 0.85 | 4.31 | 4.15 | 17.71 | 2.97 | 90 | |
| SUBARU-singleDevice=Desktop, Inference time (ms)=1.742025.06 | 0.86 | 4.19 | 4.17 | 17.49 | 2.97 | 90 | |
| SUBARU-singleDevice=Pixel7, Inference time (ms)=70.412025.06 | 0.86 | 4.14 | 4.14 | 17.31 | 2.95 | 90 | |
| HiFi++Device=Desktop, Inference time (ms)=6.32025.06 | 0.89 | 4.18 | 4.11 | 17.48 | 2.85 | 90 | |
| HiFi++Device=Pixel7, Inference time (ms)=2582025.06 | 0.91 | 4.15 | 4.09 | 17.43 | 2.83 | 90 | |
| AERODevice=Desktop, Inference time (ms)=362025.06 | 0.97 | 4.16 | 4.03 | 17.03 | 2.93 | 89 | |
| AERODevice=Pixel7, Inference time (ms)=14412025.06 | 0.98 | 4.14 | 4.01 | 16.97 | 2.89 | 88 | |
| EBENDevice=Desktop, Inference time (ms)=12.72025.06 | 1.15 | 3.78 | 3.65 | 14.23 | 2.57 | 84 | |
| EBENDevice=Pixel7, Inference time (ms)=5202025.06 | 1.17 | 3.76 | 3.63 | 14.19 | 2.56 | 84 | |
| SEANetDevice=Desktop, Inference time (ms)=9.182025.06 | 1.39 | 3.89 | 3.78 | 14.31 | 2.43 | 89 | |
| SEANetDevice=Pixel7, Inference time (ms)=3802025.06 | 1.4 | 3.85 | 3.75 | 14.24 | 2.4 | 88 | |
| TFiLMDevice=Desktop, Inference time (ms)=4.852025.06 | 1.68 | 3.73 | 3.53 | 10.28 | 2.03 | 81 | |
| TFiLMDevice=Pixel7, Inference time (ms)=1982025.06 | 1.69 | 3.7 | 3.5 | 10.19 | 2 | 81 | |
| UnprocessedDevice=Pixel7, Inference time (ms)=2.792025.06 | 1.83 | 1.26 | 8.52 | 1.1 | — | 70 | |
| UnprocessedDevice=Desktop, Inference time (ms)=2.782025.06 | 1.84 | 1.27 | 8.54 | 1.11 | — | 71 | |
| VibVoiceDevice=Desktop, Inference time (ms)=17.22025.06 | 3.1 | 2.73 | 2.51 | 12.48 | 2.05 | 81 | |
| VibVoiceDevice=Pixel7, Inference time (ms)=7072025.06 | 3.11 | 2.7 | 2.47 | 12.45 | 2.01 | 80 |