“ASR Status Report 2017-2-27”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
 
第40行: 第40行:
 
|Ying Shi   
 
|Ying Shi   
 
||  
 
||  
*  
+
* speech speaker gradinet visualization(DNN model)[http://192.168.0.51:5555/cgi-bin/cvss/cvss_request.pl?account=shiying&step=view_request&cvssid=599 here]
 +
* RNN baseline about speech recognition and speaker recognition
 +
** LSTM: performance is so bad
 +
** GRU:39.74%(a small test only 100 utterence)
 
||  
 
||  
*  
+
* speech baseline and joint baseline on fisher
 
|-
 
|-
  

2017年2月28日 (二) 05:08的最后版本


Date People Last Week This Week
2017.2.27


Jingyi Lin
Yanqing Wang
Hang Luo
  • Run 3-layer joint training network, do some visualization work and try residual network.

[1]

  • Compared ASR model in joint traning model with ASR model given fixed LID projection.
Ying Shi
  • speech speaker gradinet visualization(DNN model)here
  • RNN baseline about speech recognition and speaker recognition
    • LSTM: performance is so bad
    • GRU:39.74%(a small test only 100 utterence)
  • speech baseline and joint baseline on fisher
Yixiang Chen
  • Submit results of ASVspoofing challenge.
  • ASVspoofing -- Waiting for the key files of eva set.
Lantian Li
  • Submit results of ASVspoofing challenge [2].
  • Write book -- Done.
  • Deep speaker embedding.
Zhiyuan Tang
  • explore sub-questions,baselines
  • poster for ICASSP17
  • continue exploring sub-questions

Date People Last Week This Week
2017.2.20


Jingyi Lin
Yanqing Wang
Hang Luo
Ying Shi
Yixiang Chen
  • ASVspoofing challenge.
  • ASVspoofing.
Lantian Li
  • Deep speaker embedding [3]
  • ASVspoofing [4]
  • Write book.
  • ASVspoofing
Zhiyuan Tang
  • joint training of speech and language recognition, two languages for preliminary exploration,baselines
  • explore sub-questions