“Xuewei Zhang 2016-03-07”版本间的差异
来自cslt Wiki
| 第1行： | 第1行： | ||
* Thuyg20 | * Thuyg20 | ||
| − | :*sumit thuyg20- | + | :*sumit thuyg20-openslr |
:* thuyg recipe | :* thuyg recipe | ||
::*noise training using 0db training and test data | ::*noise training using 0db training and test data | ||
2016年3月7日 (一) 00:46的最后版本
- Thuyg20
- sumit thuyg20-openslr
- thuyg recipe
- noise training using 0db training and test data
- thuyg_sre recipe
- reproduce experiment
- 20 of frame length performs better
- clean ubm,train PLDA with noise. The performance is worse.
- mix ubm, train PLDA
