Automatic Speech Recognition on DIRHA (test)
0.298WERPASE-FineTuned
Evaluation Results
| Method | Links | |
|---|---|---|
| PASE-FineTunedClassifier=GRU, Mode=Fine-Tuned, Pre-training=Clean LibriSpeech data2019.04 | 0.298 | |
| PASE-FrozenClassifier=GRU, Mode=Frozen, Pre-training=Clean LibriSpeech data2019.04 | 0.325 | |
| PASE-SupervisedClassifier=GRU, Training=Supervised from scratch2019.04 | 0.335 | |
| FBANKClassifier=GRU2019.04 | 0.34 | |
| MFCCClassifier=GRU2019.04 | 0.358 | |
| SincNetInput features=Raw waveform, initialization=average of 5 seeds2018.11 | 0.372 | |
| CNNInput features=Raw waveform, initialization=average of 5 seeds2018.11 | 0.4 | |
| CNNInput features=FBANK, initialization=average of 5 seeds2018.11 | 0.401 |