A package to turn scatter plots into perceptually uniform sound files for use in science and to make science more accessible. This package should not be confused with SoniPy a sonification python module package by David Warrall, which can be found here.
This project was developed by Locke Patton and Prof. Emily Levesque. Click here for a twitter thread explaining the motivation behind this project.
Here is an example sonification.
Our method of sonification takes in scatterplot data and produces audio clips that depict each datapoint as a short sound blip with a y value corresponding to pitch and an x value corresponding to arrival time.
Each data point has a corresponding short tone called a blip, with a y value corresponding to its pitch and a x value corresponding to its arrival time. Higher y value data points have higher corresponding blip pitches.
soni-py is pip-installable from command line, as follows:
pip install sonipyAlternately, you can clone the repository and install it yourself, also in command line:
git clone https://github.com/lockepatton/sonipy.git
cd sonipy
python setup.py installFor two arrays of the same length, called x and y, you can sonify them using the following:
from sonipy.sonify import SonifyTool
Tone = SonifyTool(x, y)
Tone.play()
Tone.save()If you would like more fine control of the sonification inputs, you can adjust the underlying arguments as follows. For details about the parameters involved, see the Parameter Inputs section below.
from sonipy.sonify import SonifyTool
C4 = 261.6 # Hz
frequency_args = {
'frequency_min' : C4,
'frequency_max' : C4*4
# 'cents_per_value' : -680,
# 'value_min' : 0,
# 'value_max' : 1,
}
duration_args = {
'time_total' : 2000, # ms
# 'time_min' : 100, # ms
# 'time_max' : 800, # ms
}
duration_scale = 1. / 2000. # x value / time (ms)
Tone = SonifyTool(x, y,
frequency_args = frequency_args,
duration_args = duration_args,
# duration_scale = duration_scale,
bliplength=0.5)
Tone.play()
Tone.SaveTone()All frequency parameters are entered inside the frequency_args parameter. The following inputs are all accepted.
- a minimum frequency
and it's corresponding minimum y value
- a maximum frequency
and it's corresponding maximum y value
- a change in pitch (measured in cents) over change in y value parameter
By default, the sound files are 2 seconds. Time parameters are entered by simply by defining a duration_scale (in seconds per x value). Or alternately by passing a duration_args dictionary with some total time, smallest delta time between points or max delta time between points.
- a total time of the soundfile
- a change in time (measured in seconds) over change in x value parameter
Several Jupyter notebook demos that demonstrate some use cases and examples of sonipy are found here, with an interactive version found here.
This code was developed as part of TransientZoo, a citizen science program that will allow participants, including blind and visually impaired individuals, to classify supernova lightcurves using sound. In astronomy, lightcurves depict variations in brightness of a specific astrophysical object as a function of time. For more, see this summary twitter thread and poster from the 235th American Astronomical Meeting.
Thank you to Prof. Allen Downey for permission to host his thinkDSP code in this repository for easier distribution. This work wouldn't be possible without it. For more details about his book Think DSP: Digital Signal Processing in Python, see his textbook repository at https://github.com/AllenDowney/ThinkDSP.
Work on this project is welcomed. For more information on contributing, see our contributing.md guidelines.
Have an issue with your operating system or any other suggestions/improvements? Let us know by opening an issue! Have a suggestion for how to make this code more accessible? Send Locke an email at locke.patton@cfa.harvard.edu or reach out via a github issue.
If you find the package useful in your research, please cite our JOSS paper.