“ASR Status Report 2016-11-21”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
第65行: 第65行:
 
||  
 
||  
 
* training with shared layers;
 
* training with shared layers;
*  
+
* explore single tasks.
 
|}
 
|}
  

2016年11月21日 (一) 00:52的版本

Date People Last Week This Week


2016.11.21 Hang Luo
Ying Shi
  • fighting with kazak speech recognition system:because the huge size of HCLG.fst the decoding job always make the sever done.

There are several method I have tried

  • change the size or word list and corpus this method not worked very well
  • prune the LM .And the parameter been used to prune the LM is 2e-7 the size of LM reduce from 290M to 60M but the result about wer is very poor
  • I have upload some result about several experiment to CVSS[1]
  • there are too much private affairs about myself so the job about visualization last week has been delayed I will try my best to finish it the week



Yixiang Chen
  • Learn MFCC extraction mechanism.
  • Read kaldi computer-feature code and find how to change MFCC.
  • Frequency-weighting based feature extraction.
  • Continue replay detection (Freq-Weighting and Freq-Warping).
Lantian Li
  • Joint-training on SRE and LRE (LRE task). [2]
    • Tdnn is better than LSTM.
    • LRE is a long-term task.
  • Briefly overview Interspeech SRE-related papers.
  • CSLT-Replay detection.
    • Baseline done (Freq / Mel domain).
    • performance-driven based Freq-Weighting and Freq-Warping --> Yixiang.
  • LRE task.
  • Replay detection.
Zhiyuan Tang
  • report for Weekly Reading (a brief review of interspeech16), just prepared;
  • language scores as decoding mask (1.multiply probability, very bad; 2.add log-softmax, a little bad)
  • training with mask failed
  • training with shared layers;
  • explore single tasks.




Date People Last Week This Week


2016.11.14 Hang Luo
  • read papers about highway connection and multi-task
  • Explore the language recognition model on speech+language joint training, find how to use languange information.
  • finish ML-book
Ying Shi
  • kazaka recognition baseline finished here
  • figuie of ml-book read paper nn visualization
Yixiang Chen
  • Motify the ML-book and read paper.
  • Prepare the replay detection baseline.[3]
  • Complete the replay baseline and attempt to modify MFCC calculation.
Lantian Li
  • Complete the Joint-training on TASLP (speaker parts). [4]
  • Joint-training on SRE and LRE (Still over-fitting !). [5]
  • Read some papers and download four database. [6]
  • CSLT-Replay detection database is OK! [/work4/lilt/Replay]
  • Joint-training on SRE and LRE.
  • Baseline system on replay detection.
Zhiyuan Tang
  • finished the additinal experiments of joint learning (speech & spk) for taslp (multi-task, ivector as part of input)[7].
  • prepare a brief review of interspeech16.
  • report for Weekly Reading (a brief review of interspeech16);
  • joint training for bilingual: language scores as decoding mask, explore the best info receivier by studying single tasks with extra info.