CITATION — REFERENCE ENTRY
Simultaneous Estimation of Chords and Musical Context From Audio — IEEE Transactions on Audio, Speech, and Language Processing
- Key
- mauch2010-chords-musical-context
- Authors
- Mauch, Matthias; Dixon, Simon
- Issued
- 2010-8
- Type
- article-journal
- Container
- IEEE Transactions on Audio, Speech, and Language Processing
- Volume
- 18
- Issue
- 6
- Pages
- 1280-1289
Raw CSL JSON
{
"DOI": "10.1109/TASL.2009.2032947",
"URL": "https://webspace.eecs.qmul.ac.uk/s.e.dixon/pub/2010/Mauch-Dixon-TASLP-2010-real.pdf",
"page": "1280-1289",
"type": "article-journal",
"issue": "6",
"title": "Simultaneous Estimation of Chords and Musical Context From Audio",
"author": [
{
"given": "Matthias",
"family": "Mauch"
},
{
"given": "Simon",
"family": "Dixon"
}
],
"issued": {
"date-parts": [
[
2010,
8
]
]
},
"volume": "18",
"container-title": "IEEE Transactions on Audio, Speech, and Language Processing"
}
Claims
-
Each additional musical context variable improved performance, with bass modelling giving the greatest gain.
"We compared ten different variants of our algorithm and show that each additional musical parameter significantly improves the method's performance. The greatest enhancement is achieved by additional bass modeling."
-
Mauch and Dixon's six-layer dynamic Bayesian network jointly models metric position, key, chord and bass pitch class from bass and treble audio features.
"The core of the method is a six-layered dynamic Bayesian network, in which the four hidden source layers jointly model metric position, key, chord, and bass pitch class, while the two observed layers model low-level audio features corresponding to bass and treble tonal content."
-
The model's parameters encoded expert musical knowledge rather than being learned from a dataset.
"we choose to map expert musical knowledge onto a probabilistic framework, rather than learning parameters from a specific data set."
-
Chords in first inversion were recognized as such only 15% of the time.
"Chords in first inversion are recognized as such only 15% of the time"
-
The method classified 71% of chords correctly on the 176 tracks of the MIREX 2008 chord detection task.
"We show that with 71% correctly classified chords our method significantly exceeds the state of the art when tested against manually annotated ground truth transcriptions on the 176 audio tracks from the MIREX 2008 Chord Detection Task."
Available in