MUSAN: A Music, Speech, and Noise Corpus

MUSAN is a corpus of music, speech, and noise recordings designed for training models for voice activity detection and music/speech discrimination. This is a comprehensive collection suitable for various audio processing tasks.

Dataset Structure

The dataset is organized into three main categories:

1. Music (~42 hours)

• Subcategories: Classical, Pop/Rock, Jazz, and more
• Sources: Free Music Archive, Jamendo, and others
• Files: Contains instrumental and vocal music

2. Speech (~60 hours)

• Sources: LibriVox (public domain audiobooks), US Government recordings
• Languages: Primarily English
• Content: Various speakers reading books and government proceedings

3. Noise (~7 hours)

• Types: Environmental sounds, technical noises, etc.
• Source: Free Sound and sound effect collections
• Use Cases: Background noise for robust model training

Citation

If you use this dataset, please cite the original paper:

@misc{musan2015,
  author = {David Snyder and Guoguo Chen and Daniel Povey},
  title = {{MUSAN}: {A} {M}usic, {S}peech, and {N}oise {C}orpus},
  year = {2015},
  eprint = {1510.08484},
  note = {arXiv:1510.08484v1}
}

Acknowledgments

This dataset was created by David Snyder, Guoguo Chen, and Daniel Povey with support from:

• National Science Foundation Graduate Research Fellowship (Grant No. 1232825)
• Spoken Communications

License Information

All content in this corpus is available under Creative Commons licenses or is in the USA Public Domain. Each subdirectory contains:

• LICENSE file with specific licensing details for files in that directory
• ANNOTATIONS file with metadata about the recordings

For more information about Creative Commons licenses, visit: http://creativecommons.org/licenses/

Contact

Original dataset curator: David Snyder ([email protected])

---

This Hugging Face version is provided by FluidInference to make the dataset more accessible for the research community.