English

Learning Disentangled Audio Representations through Controlled Synthesis

Sound 2024-02-19 v1 Machine Learning Audio and Speech Processing

Abstract

This paper tackles the scarcity of benchmarking data in disentangled auditory representation learning. We introduce SynTone, a synthetic dataset with explicit ground truth explanatory factors for evaluating disentanglement techniques. Benchmarking state-of-the-art methods on SynTone highlights its utility for method evaluation. Our results underscore strengths and limitations in audio disentanglement, motivating future research.

Keywords

Cite

@article{arxiv.2402.10547,
  title  = {Learning Disentangled Audio Representations through Controlled Synthesis},
  author = {Yusuf Brima and Ulf Krumnack and Simone Pika and Gunther Heidemann},
  journal= {arXiv preprint arXiv:2402.10547},
  year   = {2024}
}

Comments

12 pages, 12 figures, accepted as a Tiny paper at ICLR 2024

R2 v1 2026-06-28T14:50:30.738Z