Difference between revisions of "2020-05-25"

From cslt Wiki
Jump to: navigation, search
 
(4 intermediate revisions by 4 users not shown)
Line 19: Line 19:
 
|Yunqi Cai
 
|Yunqi Cai
 
||   
 
||   
*  
+
* experiments on DSC flow ivector
 +
* experiments on DSC flow ivector constant logdet
 +
* investigate constant logdet NF flow
 
||
 
||
 
*  
 
*  
Line 30: Line 32:
 
|Zhiyuan Tang
 
|Zhiyuan Tang
 
||  
 
||  
*
+
* DNF for asr.
 
||
 
||
*
+
* Continue.
 
||
 
||
 
*   
 
*   
Line 41: Line 43:
 
|Lantian Li
 
|Lantian Li
 
||  
 
||  
*
+
* Enroll-test mismatch (data)
 +
* DT-DNF for ASV.
 
||
 
||
*
+
* Enroll-test mismatch (NL scoring)
 +
* DT-DNF for ASV.
 
||
 
||
 
*   
 
*   
Line 52: Line 56:
 
|Ying Shi
 
|Ying Shi
 
||  
 
||  
*
+
* compute the result about flow denoise and glow denoise on SDR  PESQ fwSNR
 +
* train new flow with both postive samples and negative samples
 +
* train RPCA baseline
 
||
 
||
*  
+
* finish the result form with RPCA and the new flow
 +
* train an energy model
 
||
 
||
 
*   
 
*   
Line 73: Line 80:
 
|Yue Fan
 
|Yue Fan
 
||  
 
||  
*  
+
* Check cn2
 +
* Gun data preparation
 +
* Do exprintments on mdl-cn
 
||
 
||
*
+
* Perform more experiments on gunshot recognition with speaker recognition
 
||
 
||
 
*   
 
*   

Latest revision as of 01:03, 25 May 2020

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Some experiments with flow-based denoising: GMM-based noise model, spectrum flow criterion.
  • Some literature reivew with deep denoising methods
  • More literature reivew for denosing
  • More literature review/theory design for NDA.
Yunqi Cai
  • experiments on DSC flow ivector
  • experiments on DSC flow ivector constant logdet
  • investigate constant logdet NF flow
Zhiyuan Tang
  • DNF for asr.
  • Continue.
Lantian Li
  • Enroll-test mismatch (data)
  • DT-DNF for ASV.
  • Enroll-test mismatch (NL scoring)
  • DT-DNF for ASV.
Ying Shi
  • compute the result about flow denoise and glow denoise on SDR PESQ fwSNR
  • train new flow with both postive samples and negative samples
  • train RPCA baseline
  • finish the result form with RPCA and the new flow
  • train an energy model
Haoran Sun
Yue Fan
  • Check cn2
  • Gun data preparation
  • Do exprintments on mdl-cn
  • Perform more experiments on gunshot recognition with speaker recognition
Jiawen Kang
  • Organize meta-learning code
  • Check cn2
  • Prepare shared PPT
  • Prepare cross-channel and near-far data.
Ruiqi Liu
  • Kaldi baseline experiments on cn2.
  • Get some information about speakers to analyze cn2.
  • Other experiments.
Sitong Cheng
Zhixin Liu
Haolin Chen