English

Shaking Acoustic Spectral Sub-bands Can Better Regularize Learning in Affective Computing

Sound 2018-04-19 v1 Audio and Speech Processing

Abstract

In this work, we investigate a recently proposed regularization technique based on multi-branch architectures, called Shake-Shake regularization, for the task of speech emotion recognition. In addition, we also propose variants to incorporate domain knowledge into model configurations. The experimental results demonstrate: 1)1) independently shaking sub-bands delivers favorable models compared to shaking the entire spectral-temporal feature maps. 2)2) with proper patience in early stopping, the proposed models can simultaneously outperform the baseline and maintain a smaller performance gap between training and validation.

Keywords

Cite

@article{arxiv.1804.06779,
  title  = {Shaking Acoustic Spectral Sub-bands Can Better Regularize Learning in Affective Computing},
  author = {Che-Wei Huang and Shrikanth Narayanan},
  journal= {arXiv preprint arXiv:1804.06779},
  year   = {2018}
}

Comments

ICASSP paper with follow-up exps

R2 v1 2026-06-23T01:27:44.682Z