2015年7月28日 (二) 10:45的版本

Task To Do

• End-to-End speech recognition

• Zhiyuan Tang/Mengyuan Zhao/Zhiyong Zhang

• Real environment noise cancellation

• (RNN-DAE: echo or reverberation)
• Mengyuan Zhao/Zhiyong Zhang

• Integrate the class information to HCLG fst for speech recognition

• Distant speech recognition

• (Reverberation, Mutli-microphones)*
• (Lasso),Xuewei Zhang

• Voice conversation

• xx

• Unbound activation function(Rectifier/Maxout/Pnorm) go-through searching method

• Zhiyong Zhang

• Sparse DNN

• Zhiyuan Tang

• DNN training GPU parallelization , nnet2 optimization

• Monmentum-like Hessien-Free acceleration

• Correlation based SEONE cluster

• NN GPU parallel traing

• Sheng Su

• Audio Embedding

• Activation value normalization through time

• For bigger learning rate

• Mix-training Balance decision tree

• Zhiyong Zhang

• RNN training accelerating

• Data selection

• Zhiyong Zhang
• Sub-modular data selection

• Decoder

• Confidence output for task-required

Task DONE

• Multi-Mode features based VAD*

• Shi Yin, DONE

• DNN based Language identification and Speaker identification*

• Xuewei Zhang/Zhiyuan Tang

• Neural network visulization*

• Mian Wang,DONE

• Dark knowledge*

• Mengyuan Zhao, Xiangyu Zeng, Zhiyong Zhang, Chao Liu

• Normal RNN speech recognition*

• Mengyuan Zhao

Technical Report To Write

1, DNN-DAE based noise cancellation -- Xiangyu Zeng / Mengyuan Zhao / Zhiyong Zhang  --DONE
2, Speech Rate DNN speech recognition --Shi Yin/Xiangyu Zeng --DONE
3, CNN+fbank feature combination --Mian Wang /Yiye Lin /Mengyuan Zhao /Shi Yin
4, Uyghur low-resource acoustic model enhancement -- Shi Yin / Mengyuan Zhao / Zhiyong Zhang --DONE
5, Uyghur 20h database release --Kaer /Shi Yin --DONE
6，Dark-Knowledge Transfer
   *: Xiangyu Zeng/ Mengyuan Zhao / Zhiyong Zhang

Paper to Write

Project

• Xiaomi TV

• Mengyuan Zhao/Zhiyong Zhang
• TAG-lm & Domain-specific general lm

• Chinese-English mix-training