This repository contains tools for studying rhythms in symbolic format. It was developed for the study of polyphonic drum patterns, but can be adapted for other types of patterns. It implements various descriptors derived from scientific literature; see Descriptors for a full list with references.
pip install git+https://github.com/danielgomezmarin/rhythmtoolbox
Rhythm Toolbox supports multiple representations of symbolic rhythm. Across all representations, Rhythm Toolbox operates at a 16th note resolution, or 4 ticks per beat in MIDI terms. If data is passed in at a different resolution, it is resampled by associating each onset with its closest 16th note position.
To compute descriptors from a MIDI file:
from rhythmtoolbox import midifile2descriptors
midifile2descriptors('midi/boska/3.mid')A piano roll is a (N, V) matrix, where N
is a number of time steps and V is a number of MIDI pitches. Any positive value represents an onset; Rhythm Toolbox
does not currently consider velocity.
To compute descriptors from a piano roll:
from rhythmtoolbox import pianoroll2descriptors
pianoroll2descriptors(roll)A pattern list is a list of lists representing time steps, each containing the MIDI note numbers that occur at that step.
To compute descriptors from a pattern list:
from rhythmtoolbox import pattlist2descriptors
pattlist = [
[36, 38, 42],
[],
[],
[38, 42],
[46],
[46],
[36, 38, 42],
[],
[42],
[38],
[36, 42],
[],
[38, 46],
[46],
[42, 64],
[],
]
pattlist2descriptors(pattlist)The following descriptors are discussed in Gómez-Marín et al, 2020. Additional sources are listed where applicable. The mapping of MIDI instruments to frequency bands can be found in midi_mapping.py.
| Name | Description |
|---|---|
| noi | Number of instruments |
| polyDensity | Number of onsets |
| lowDensity | Number of onsets in the low freq band |
| midDensity | Number of onsets in the mid freq band |
| hiDensity | Number of onsets in the high freq band |
| lowness | Fraction of onsets in the low freq band |
| midness | Fraction of onsets in the mid freq band |
| hiness | Fraction of onsets in the high freq band |
| stepDensity | Fraction of steps with onsets |
The following descriptors are valid only for 16-step patterns:
| Name | Description | Reference |
|---|---|---|
| sync | Syncopation | |
| lowSync | Syncopation of the low freq band | |
| midSync | Syncopation of the mid freq band | |
| hiSync | Syncopation of the high freq band | |
| syness | Syncopation / density | |
| lowSyness | Syncopation / density of the low freq band | |
| midSyness | Syncopation / density of the mid freq band | |
| hiSyness | Syncopation / density of the high freq band | |
| balance | Monophonic balance | Milne and Herff, 2020 |
| polyBalance | Polyphonic balance | |
| evenness | Monophonic evenness | Milne and Dean, 2016 |
| polyEvenness | Polyphonic evenness | |
| polySync | Polyphonic syncopation | Witek et al, 2014 |
If this repository is useful for your research please cite the following paper via the BibTeX below.
@article{gomez2020drum,
title={Drum rhythm spaces: From polyphonic similarity to generative maps},
author={G{\'o}mez-Mar{\'\i}n, Daniel and Jord{\`a}, Sergi and Herrera, Perfecto},
journal={Journal of New Music Research},
volume={49},
number={5},
pages={438--456},
year={2020},
publisher={Taylor \& Francis}
}
To cite the initial version of this repository (Nov 2018), checkout
commit 6acdb69a60153d08.
The MIDI drum patterns examples included in midi/boska and midi/sano were provided by Jon-Eirik Boska and Sebastián Hoyos, respectively.