13-03-16 Mengyuan Zhao
From cslt Wiki
- Compare Fbank and GFbank feature.
- Amr adaptation:
- trained xEnt and mpe, compare their perfomance.
- Convert 1700H wav into amr then convert back.
- Generate Fbank feature of 1700H corpus(amr).
- Trained gmm model on 1700H data.
Target for next week:
- Combine xEnt and mpe, test it.
- Fix the problem in 1700H gmm training.
- do mpe on sinovoice 1700H model using amr data(1700H)