Skip to content

glam-imperial/NSL-SER

Repository files navigation

Knowledge Transfer For On-Device Speech Emotion Recognition with Neural Structured Learning

arXiv License: MIT

NSL-SER

This is the code for the proposed neural structured learning (NSL) framework for speech emotion recognition (SER) in the following paper:

Yi Chang, Zhao Ren, Thanh Tam Nguyen, Kun Qian, and Björn W. Schuller. Knowledge Transfer For On-Device Speech Emotion Recognition with Neural Structured Learning. https://arxiv.org/abs/2210.14977

Citation

@misc{ychang2022ser,
      title={Knowledge Transfer For On-Device Speech Emotion Recognition with Neural Structured Learning}, 
      author={Chang, Yi and Ren, Zhao and Nguyen, Thanh Tam and Qian, Kun and Schuller, Björn W.},
      year={2022},
      eprint={2210.14977},
      archivePrefix={arXiv},
      note={5 pages}
}

Abstract

We propose a neural structured learning (NSL) framework through building synthesized graphs. An SER model is trained on a source dataset and used to build graphs on a target dataset. A lightweight model is then trained with the speech samples and graphs together as the input. Our experiments demonstrate that training a lightweight SER model on the target dataset with speech samples and graphs can not only produce small SER models, but also enhance the model performance over models with speech samples only.

Experiments Running

Both python files can be run via

sh run.sh

Please note that the folder /neural-structured-learning can be forked from official one: https://github.com/tensorflow/neural-structured-learning. Only the file 'graph_regularization.py' is adjusted herein.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published