From cslt Wiki
Jump to: navigation, search

Speech Processing

AM development

Sparse DNN

  • Performance improvement found when pruned slightly
  • Experiments show that
  • Suggest to use TIMIT / AURORA 4 for training

Noise training

  • First draft of the noisy training journal paper
  • Check abnormal behavior with large sigma (Yinshi, Liuchao)

Drop out & Rectification & convolutive network

  • Drop out
  • No performance improvement found yet.
  • [1]
  • Rectification
  • Dropout NA problem was caused by large magnitude of weights
  • Convolutive network
  1. Test more configurations
  • Zhiyong will work on CNN
  • Recurrent neural network
  • investigate CURRENNT for AM

Denoising & Farfield ASR

  • Lasso-based de-reverberation is done with the REVERBERATION toolkit
  • Start to compose the experiment section for the SL paper.


  • problems found at the beginning part of speech (0-0.02s?)
  • Noise model training done. Under testing.
  • Need to investigate the performance reduction in babble noise. Call Jia.

Speech rate training

  • Some interesting results with the simple speech rate change algorithm was obtained on the WSJ db


  • Seems ROS model is superior to the normal one with faster speech
  • Need to check distribution of ROS on WSJ
  • Suggest to extract speech data of different ROS, construct a new test set
  • Suggest to use Tencent training data
  • Suggest to remove silence when compute ROS

low resource language AM training

  • Results on CVSS[3]
  • Use Chinese NN as initial NN, change the last layer


  • global scoring done.
  • Pitch & rhythm done, need testing
  • Harmonics hold


  • experiments done, need more data
  • Basic confidence by using lattice-based posterior + DNN posterior + ROS done
  • 23% detection error achieved by balanced model

Speaker ID

  • Add VAD to system
  • GMM-based test program delivered
  • GMM registration program done

Emotion detection

  • Zhang Weiwei is learning the code
  • Sinovoice is implementing the server

Text Processing

LM development

Domain specific LM

h2. ngram generation is on going h2. look the memory and baidu_hi done

h2. NUM tag LM:

  • maxi work is released.
  • yuanbin continue the tag lm work.
  • add the ner to tag lm .
  • Boost specific words like wifi if TAG model does not work for a particular word.


W2V based doc classification

  • Initial results variable Bayesian GMM obtained. Performance is not as good as the conventional GMM.
  • Non-linear inter-language transform: English-Spanish-Czch: wv model training done, transform model on investigation
  • SSA-based local linear mapping still on running.
  • k-means classes change to 2.
  • Knowledge vector started
  • document obtained from wiki
  • formula obtained
  • Character to word conversion
  • read more paper .
  • prepare to train .
  • Google word vector train
  • improve the sampling method


  • Prepare WSJ database
  • Trained model 10000 x 4 + 320 + 10000
  • Better performance obtained (4.16-3.47)
  • gigaword sampling for Chinese data


  • v3.0 demo released
  • still slow
  • cut the vocabulary that is not important .


  • liangshan_v1 performance 74.3%.
  • New framework and GA method is done
  • add SEMPRE tool to framework