CITATION — REFERENCE ENTRY

Chord Recognition: Evaluation — FMP Notebooks

Revision 0660190b-b665-4126-a881-69d06f98aaf9 · 11/10/2026, 11:56:16 PM UTC
Key
fmp-c5-chord-evaluation
Authors
Müller, Meinard; Weiß, Christof
Type
webpage
Container
FMP Notebooks
Publisher
International Audio Laboratories Erlangen
Raw CSL JSON
{
  "URL": "https://www.audiolabs-erlangen.de/resources/MIR/FMP/C5/C5S2_ChordRec_Eval.html",
  "type": "webpage",
  "title": "Chord Recognition: Evaluation",
  "author": [
    {
      "given": "Meinard",
      "family": "Müller"
    },
    {
      "given": "Christof",
      "family": "Weiß"
    }
  ],
  "accessed": {
    "date-parts": [
      [
        2026,
        11,
        10
      ]
    ]
  },
  "publisher": "International Audio Laboratories Erlangen",
  "container-title": "FMP Notebooks"
}

Claims

  1. A single played note can produce substantial energy in the chroma bands of its harmonics, such as the fifth (third harmonic) and the major third (fifth harmonic), which often leads to confusion between major and minor chords.
    "This often results in some problem known as major–minor confusion."
    Locator: section: Problems · Quote language: en
  2. In general no well-defined ground truth exists for chord recognition, and even music experts may disagree on how to annotate a piece.
    "In general, however, no well-defined ground truth exists, and even music experts may disagree on how to annotate a given piece of music."
    Locator: section: Evaluation · Quote language: en
  3. Many recognition errors stem from oversimplified chord models such as the 24 major and minor triads, but enlarging the vocabulary also increases the chance of confusion between chords.
    "However, on the downside, increasing the number of possible chords also increases the confusion probability in the classification stage."
    Locator: section: Problems · Quote language: en
Available in