Empirical Evaluation of Speaker Adaptation on DNN based Acoustic Model

Introduction

This is the repository of the acoustic model adaptation. Our original paper can be found here.

In this work we give an empirical evaluation of three typical speaker adaptation methods: Linear Input Network (LIN), Linearing Hidden Unit Contribution (LHUC) and Kullback-Leibler divergence (KLD) regularization.

Dependencies

• kaldi-wangke

Reference

If the code of this repository was useful for your research, please cite our work:

@inproceedings{Wang2018Empirical,
  address   = {Hyderabad, India},
  author    = {Wang, Ke and Zhang, Junbo and Wang, Yujun and Xie, Lei},
  booktitle = {Proceedings of Interspeech},
  pages     = {2429--2433},
  publisher = {ISCA},
  title     = {{Empirical Evaluation of Speaker Adaptation on DNN based Acoustic Model}},
  url       = {https://www.isca-speech.org/archive/Interspeech{\_}2018/pdfs/1897.pdf},
  year      = {2018}
}

Contact

e-mail: [email protected]