Xiangyu Zeng 2015-03-30

From cslt Wiki
Jump to: navigation, search

last week:

1.accomplished some experiments of ROS: sampled half, slow-based, fast-based data set to train DNN and ROS compensation structure

2.write a script to add ROS feature into GMM model and rewrite the script of decode gmm.

3.completed the paper of InterSpeech

this week:

1.change the inserting position of ROS feature to avoid some influence of LDA.

2.begin to write the technical report of DAE