Metric Modulations Dataset

We present a dataset of annotations of tracks containing metric modulations.

A metric modulation is defined by a relatively abrupt change of metrical structure over time.
In order to capture this, for each track the annotations consist of:

  • Beat locations
  • Downbeat locations
  • Metric modulation locations (i.e. timestamps of the metrical structure change)
  • All metrical level pulse rates

It comes that the metrical structure (i.e. beat rate, downbeat rate and all other metrical level pulse rates) are relatively consistent between two metric modulations.

Download the dataset here

Dataset Contents:

  • READ_ME.txt: contains useful information about the dataset, such as a description of the data structure, citation information etc.
  • MetModAnn_Published.json: contains the annotations for all the tracks

If you make use of this dataset, please cite:
[1] E. Quinton, K. O'Hanlon, S. Dixon and M. Sandler, “Tracking Metrical Structure Changes With Sparse NMF,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2017.
Read paper here

The audio recordings making up this dataset are under copyright and can therefore not be publicly shared here.
However, the annotation file specifies the usual track information with the addition of the ISRC, which uniquely identifies the very recordings that have been used so that they can be retrieved from any suitable platform (What is an ISRC?).
We are also happy to help interested researchers to access audio data.

Get in touch with us: