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: Beyond chord vocabularies: Exploiting pitch-relationships in a chord estimation metric

Abstract: Chord estimation metrics treat chord labels as independent of one another. This fails to represent the pitch relationships between the chords in a meaningful way, resulting in evaluations that must make compromises with complex chord vocabularies and that often require time-consuming qualitative analyses to determine details about how a chord estimation algorithm performs. This paper presents an accuracy metric for chord estimation that compares the pitch content of the estimated chords against the ground truth that captures both the correct notes that are estimated and additional notes that are inserted into the estimate. This is not a stand-alone evaluation protocol but rather a metric that can be integrated as a weighting into existing evaluation approaches.
Comments: Extended abstract, 3 pages, 2 tables
Subjects: Sound (cs.SD); Audio and Speech Processing (eess.AS)
Journal reference: Late-Breaking Demo Session of the 22nd International Society for Music Information Retrieval Conference (2021)
Cite as: arXiv:2201.05244 [cs.SD]
  (or arXiv:2201.05244v1 [cs.SD] for this version)

Submission history

From: Johanna Devaney [view email]
[v1] Thu, 13 Jan 2022 23:34:03 GMT (117kb)

Link back to: arXiv, form interface, contact.