We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.SD

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Sound

Title: Random Projections of Mel-Spectrograms as Low-Level Features for Automatic Music Genre Classification

Abstract: In this work, we analyse the random projections of Mel-spectrograms as low-level features for music genre classification. This approach was compared to handcrafted features, features learned using an auto-encoder and features obtained from a transfer learning setting. Tests in five different well-known, publicly available datasets show that random projections leads to results comparable to learned features and outperforms features obtained via transfer learning in a shallow learning scenario. Random projections do not require using extensive specialist knowledge and, simultaneously, requires less computational power for training than other projection-based low-level features. Therefore, they can be are a viable choice for usage in shallow learning content-based music genre classification.
Comments: Submitted to IEEE Signal Processing Letters
Subjects: Sound (cs.SD); Information Retrieval (cs.IR); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
Cite as: arXiv:1911.04660 [cs.SD]
  (or arXiv:1911.04660v1 [cs.SD] for this version)

Submission history

From: Juliano Henrique Foleiss [view email]
[v1] Tue, 12 Nov 2019 03:58:11 GMT (200kb,D)

Link back to: arXiv, form interface, contact.