“2020-03-16”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
 
（3位用户的3个中间修订版本未显示）
第6行： 第6行：
 
|Dong Wang
 
|Dong Wang
 
||  
 
||  
−
*  
+
* Finish the analysis for optimal score
 +
* Design a new approach for condition transfer
 +
* Initial experiment on subspace DNF
 
||
 
||
−
*  
+
* Compte a draft on condition transfer
 +
* More experiments on subspace DNF
 +
* More investigation on distirbutional x-vector
 
||
 
||
 
*   
 
*   
第39行： 第43行：
 
|Lantian Li
 
|Lantian Li
 
||  
 
||  
−
*  
+
* Prepare back-end scoring (python vs. Kaldi).
 +
* Fast decoding of xvector in tf-kaldi.
 +
* Clean up tools (Leven dis and local CMD).
 
||
 
||
−
*  
+
* Go on back-end scoring.
 
||
 
||
 
*   
 
*   
第124行： 第130行：
 
|Sitong Cheng
 
|Sitong Cheng
 
||  
 
||  
−
*  
+
* DNF of pronunciation scoring.
 
||
 
||
−
*  
+
* Still DNF.
 +
* Do some experiments.
 
||
 
||
 
*   
 
*   
第135行： 第142行：
 
|Zhixin Liu
 
|Zhixin Liu
 
||  
 
||  
−
*  
+
* Train baseline
 
||
 
||
−
*  
+
* Analyze result
 
||
 
||
 
*   
 
*   

2020年3月16日 (一) 11:31的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Finish the analysis for optimal score
  • Design a new approach for condition transfer
  • Initial experiment on subspace DNF
  • Compte a draft on condition transfer
  • More experiments on subspace DNF
  • More investigation on distirbutional x-vector
Yunqi Cai
  • investigate energy model
  • investigate energy model on NL socre
Zhiyuan Tang
  • Conditioned flow with waveglow backend.
  • Conditioned flow with waveglow backend.
Lantian Li
  • Prepare back-end scoring (python vs. Kaldi).
  • Fast decoding of xvector in tf-kaldi.
  • Clean up tools (Leven dis and local CMD).
  • Go on back-end scoring.
Ying Shi
  • Report about speech enhancement
  • Summarize current conclusions and make plan
  • Retrain DAE, double flow and double AE
  • Train Flow-VAE
  • Prepare paper
Wenqiang Du
Haoran Sun
Yue Fan
  • Complete half of the data collection
  • Promotion of crawler tools
  • Simultaneously work on the data download and pipeline process of cn2
  • Train dnf model with gun&nongun and test
Jiawen Kang
  • Learning Meta learning
  • Experiments report
  • Arragement code.
  • implement metalearning
Ruiqi Liu
  • Training model and arranging experiments data.
  • Arranging experiments code.
  • Investigate meta-learning and do experiments
Sitong Cheng
  • DNF of pronunciation scoring.
  • Still DNF.
  • Do some experiments.
Zhixin Liu
  • Train baseline
  • Analyze result
Haolin Chen
  • Train double glow with white-noised TIMIT, mel sprctrogram
  • Replace noisy model in double glow with VAE
  • Full evaluation using PESQ, STOI, fwsegSNR on white-noised test data
  • Summarize result
  • Prepare paper
取自“http://index.cslt.org/mediawiki/index.php?title=2020-03-16&oldid=34790”