“ASR Status Report 2017-7-10”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
(以“{| class="wikitable" !Date!!People !! Last Week !! This Week |- | rowspan="7"|2017.7.3 |Miao Zhang || * prepared for the report * did translation || * do a repor...”为内容创建页面)
 
 
(5位用户的12个中间修订版本未显示)
第2行: 第2行:
 
!Date!!People !! Last Week !! This Week
 
!Date!!People !! Last Week !! This Week
 
|-
 
|-
| rowspan="7"|2017.7.3
+
| rowspan="7"|2017.7.10
  
  
 
|Miao Zhang
 
|Miao Zhang
 
||  
 
||  
* prepared for the report
+
* A report about trivial events
* did translation
+
* Finish the test website with Tanghui
 
||  
 
||  
* do a report
+
* Read the paper of Paralinguistics
* finish the website
+
* Make a plan for recording and start to record hopefully.
* comfirm the next work direction with Tanghui and yixiang
+
 
|-
 
|-
  
第18行: 第17行:
 
|Hui Tang  
 
|Hui Tang  
 
||  
 
||  
* submitted the technical report
+
* completed the test web site [http://192.168.0.84:8091/speech/index.php web]
* finished to check the subset(happy) of our speech database 
+
* finished checking the subset of our speech databases (nearly 800 sentences)
* test the model of detecting overlapped speech
+
[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/e/e3/TRP-201700011.pdf here]
+
{|class="wikitable"
+
|dataset||FER of original speech ||FER of overlapped speech|| total
+
|-
+
|th30||19.4%||15.3%|| 17.3%
+
|-
+
|dataset||FER of original speech||FER of overlapped speech of same words|| FER of overlapped speech of diff words
+
|-
+
|real environment||19.4%||28%|| 37.8%
+
|
+
|-
+
|}
+
 
||  
 
||  
* check the subset(angry) of our database
+
* finish checking the reminder of the databases  (nearly  2500 sentences)
* finish the website with zhangmiao and yixiang
+
 
|-
 
|-
  
第42行: 第27行:
 
|Yanqing Wang
 
|Yanqing Wang
 
||  
 
||  
* use dnn & sigmoid to prune [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/4/4e/Dnn_sigmoid.pdf result]
+
* use different activation function to prune [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/0/03/Results.pdf result]
||  
+
||
* Importance of each layer; try Tanh.
+
* make the network narrow and test the former conclusions.
 +
* change source code to retrain the neural network.
 
|-
 
|-
  
第51行: 第37行:
 
|Ying Shi   
 
|Ying Shi   
 
||  
 
||  
* kazak ASR transfor learning
+
* help zheling to finish his first crawler program
* help zheling Zhang to learn Scrapy
+
* chech hazak speech data
 +
** train: 1346 utterances' WER large than 20% (utterance level WER)
 +
** test: 759 utterances' WER large than 20%(utterance level WER)
 +
* transfer learning based on th30 and wsj (performance is poor)
 
||  
 
||  
* finish kazak ASR transfor learning
+
* tools for speech data checking
* help zheling Zhang to write his first Web crawler
+
* transfer learning based on large Chinese ASR model
 
|-
 
|-
  
第62行: 第51行:
 
|Yixiang Chen   
 
|Yixiang Chen   
 
||  
 
||  
* absent.
+
* plot for “learning deep speaker features”
 
||  
 
||  
*  
+
* comprehend Paralinguistics
 +
* record voice
 
|-
 
|-
  
第71行: 第61行:
 
|Lantian Li   
 
|Lantian Li   
 
||  
 
||  
* write book.
+
* deep speaker feature
* check the process of deep speaker segmentation.
+
** segmentation is still not suitable.
* demo for speech synthesis.
+
** visualization with the t-sne seems cool.
 +
* help Zhangzy decode d-vector and re-train a new deep speaker model.
 
||  
 
||  
* deep speaker visulization and segmentation.
+
* more details of segmentation experiments.
 +
* prepare the weekly meeting.
 
|-
 
|-
  
第82行: 第74行:
 
|Zhiyuan Tang  
 
|Zhiyuan Tang  
 
||  
 
||  
* Compile auto-scoring system;  
+
* Scanned the source code of auto-scoring system;
* Chinglish for mix-asr training.
+
* A report about the research of the speech group (Thursday).  
 
||  
 
||  
* A report about the research of the speech group (Thursday);
+
* Replace the old version kaldi with new ones.
* Into the source code of auto-scoring system;
+
* Gather Part 1: 'Speech, Speech Processing and Tools' of Kaldi book for further release.
 
|-
 
|-
  
第92行: 第84行:
  
 
------------------------
 
------------------------
 
  
 
{| class="wikitable"
 
{| class="wikitable"
第135行: 第126行:
 
|Ying Shi   
 
|Ying Shi   
 
||  
 
||  
* kazak ASR transfor learning
+
* kazak ASR transfer learning
 
* help zheling Zhang to learn Scrapy
 
* help zheling Zhang to learn Scrapy
 
||  
 
||  
* finish kazak ASR transfor learning
+
* finish kazak ASR transfer learning
 
* help zheling Zhang to write his first Web crawler
 
* help zheling Zhang to write his first Web crawler
 
|-
 
|-

2017年7月17日 (一) 04:05的最后版本

Date People Last Week This Week
2017.7.10


Miao Zhang
  • A report about trivial events
  • Finish the test website with Tanghui
  • Read the paper of Paralinguistics
  • Make a plan for recording and start to record hopefully.
Hui Tang
  • completed the test web site web
  • finished checking the subset of our speech databases (nearly 800 sentences)
  • finish checking the reminder of the databases (nearly 2500 sentences)
Yanqing Wang
  • use different activation function to prune result
  • make the network narrow and test the former conclusions.
  • change source code to retrain the neural network.
Ying Shi
  • help zheling to finish his first crawler program
  • chech hazak speech data
    • train: 1346 utterances' WER large than 20% (utterance level WER)
    • test: 759 utterances' WER large than 20%(utterance level WER)
  • transfer learning based on th30 and wsj (performance is poor)
  • tools for speech data checking
  • transfer learning based on large Chinese ASR model
Yixiang Chen
  • plot for “learning deep speaker features”
  • comprehend Paralinguistics
  • record voice
Lantian Li
  • deep speaker feature
    • segmentation is still not suitable.
    • visualization with the t-sne seems cool.
  • help Zhangzy decode d-vector and re-train a new deep speaker model.
  • more details of segmentation experiments.
  • prepare the weekly meeting.
Zhiyuan Tang
  • Scanned the source code of auto-scoring system;
  • A report about the research of the speech group (Thursday).
  • Replace the old version kaldi with new ones.
  • Gather Part 1: 'Speech, Speech Processing and Tools' of Kaldi book for further release.

Date People Last Week This Week
2017.7.3


Miao Zhang
  • prepared for the report
  • did translation
  • do a report
  • finish the website
  • comfirm the next work direction with Tanghui and yixiang
Hui Tang
  • submitted the technical report
  • finished to check the subset(happy) of our speech database
  • test the model of detecting overlapped speech

here

  • check the subset(angry) of our database
  • finish the website with zhangmiao and yixiang
Yanqing Wang
  • use dnn & sigmoid to prune result
  • Importance of each layer; try Tanh.
Ying Shi
  • kazak ASR transfer learning
  • help zheling Zhang to learn Scrapy
  • finish kazak ASR transfer learning
  • help zheling Zhang to write his first Web crawler
Yixiang Chen
  • absent.
Lantian Li
  • write book.
  • check the process of deep speaker segmentation.
  • demo for speech synthesis.
  • deep speaker visulization and segmentation.
Zhiyuan Tang
  • Compile auto-scoring system;
  • Chinglish for mix-asr training.
  • A report about the research of the speech group (Thursday);
  • Into the source code of auto-scoring system;