CITATION — REFERENCE ENTRY
Beat Tracking by Dynamic Programming — Journal of New Music Research
- Key
- ellis2007-beat-tracking-dp
- Authors
- Ellis, Daniel P. W.
- Issued
- 2007-3
- Type
- article-journal
- Container
- Journal of New Music Research
- Volume
- 36
- Issue
- 1
- Pages
- 51-60
Raw CSL JSON
{
"DOI": "10.1080/09298210701653344",
"URL": "https://www.ee.columbia.edu/~dpwe/pubs/Ellis07-beattrack.pdf",
"page": "51-60",
"type": "article-journal",
"issue": "1",
"title": "Beat Tracking by Dynamic Programming",
"author": [
{
"given": "Daniel P. W.",
"family": "Ellis"
}
],
"issued": {
"date-parts": [
[
2007,
3
]
]
},
"volume": "36",
"container-title": "Journal of New Music Research"
}
Claims
-
Beat tracking derives from audio a sequence of beat instants corresponding to when a listener would tap their foot; selected instants should coincide with indicated beats such as note onsets, and should reflect a locally constant inter-beat interval.
"Beat tracking – i.e. deriving from a music audio signal a sequence of beat instants that might correspond to when a human listener would tap his foot – involves satisfying two constraints: On the one hand, the selected instants should generally correspond to moments in the audio where a beat is indicated, for instance by the onset of a note played by one of the instruments. On the other hand, the set of beats should reflect a locally-constant inter-beat-interval, since it is this regular spacing between beat times that defines musical rhythm."
-
Ellis's system estimates a global tempo, then uses dynamic programming to find beat times that fit the tempo and coincide with high onset strength; it scored just under 60% beat accuracy on MIREX-06 development data and could follow tempo changes of about ±10%.
"We describe a beat tracking system which first estimates a global tempo, uses this tempo to construct a transition cost function, then uses dynamic programming to find the best-scoring set of beat times that reflect the tempo as well as corresponding to moments of high ‘onset strength’ in a function derived from the audio. This very simple and computationally efficient procedure is shown to perform well on the MIREX-06 beat tracking training data, achieving an average beat accuracy of just under 60% on the development data. We also examine the impact of the assumption of a fixed target tempo, and show that the system is typically able to track tempo changes in a range of ±10% of the target tempo."
-
In MIREX-06 training data, listeners' taps for each excerpt clustered into two groups at slower and faster metrical levels (ratio 2 or 3), so no beat tracker can score close to 100% agreement with such ground truth.
"In each of these examples, the subject tapping data could be clustered into two groups, corresponding to slower and faster levels of the metrical hierarchy of the music, which were separated by a ratio of 2 or 3 [...] Because of the multiplicity of metrical levels reflected in the ground-truth data (as noted in section 3.2), it is not possible for any beat tracker to score close to 100% agreement with this data."
Available in