Semantic Representation Evaluation on ARCH (test)
48.96RAVDESSSpeechTokenizer
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SpeechTokenizerToken Rate=50, Q=8, Bitrate [kbps]=42025.12 | 48.96 | 97.25 | 35.91 | 55.66 | 37.9 | 30.27 | 50.99 | |
| SACodecToken Rate=75, Q=1, Bitrate [kbps]=0.752025.12 | 42.65 | 88.45 | 32.81 | 58.12 | 33.85 | 33.31 | 48.09 | |
| DACToken Rate=75, Q=2, Bitrate [kbps]=1.52025.12 | 42.36 | 68.4 | 30.23 | 58.59 | 28.65 | 31.66 | 43.32 | |
| FACodecToken Rate=80, Q=1, Bitrate [kbps]=0.82025.12 | 42.31 | 82.97 | 27.66 | 53.64 | 36.65 | 33.11 | 46.06 | |
| UniCodecTPS=752025.02 | 40.28 | 70.94 | 29.55 | 59.29 | 26 | 34.17 | — | |
| DACToken Rate=75, Q=8, Bitrate [kbps]=62025.12 | 38.89 | 72.95 | 33.31 | 59.48 | 24.4 | 30.77 | 43.3 | |
| UniCodecTPS=75, semantic stage=without2025.02 | 36.81 | 69.84 | 28.09 | 54.05 | 20.8 | 30.21 | — | |
| EncodecToken Rate=75, Q=8, Bitrate [kbps]=62025.12 | 35.07 | 49.13 | 30.97 | 42.26 | 20.15 | 26.75 | 34.06 | |
| WavTokenizerToken Rate=75, Q=1, Bitrate [kbps]=0.92025.12 | 34.38 | 62.92 | 26.89 | 53.22 | 23.4 | 28.15 | 38.16 | |
| Wavtokenizer (speech)TPS=75, domain=speech2025.02 | 32.55 | 69.57 | — | — | — | — | — | |
| EncodecToken Rate=75, Q=2, Bitrate [kbps]=1.52025.12 | 27.78 | 61.35 | 29.18 | 50.01 | 26.95 | 30.79 | 37.68 | |
| EncodecTPS=1502025.02 | 27.43 | 36.49 | 19 | 32.45 | 16.99 | 26.3 | — | |
| DACTPS=1002025.02 | 25 | 62.87 | 25.02 | 51.37 | 20.65 | 29.91 | — | |
| Wavtokenizer (music&audio)TPS=75, domain=music&audio2025.02 | — | — | 28.35 | 57.64 | 25.5 | 35.63 | — |