Datascience sonification.

[Gabriele Battaglia]

Hi everyone,
I would like to learn a tool that allows me to sonify a series of
numerical data. I'm afraid I don't have the skills to create one myself,
even though I've tried and gotten some results. However, I would like to
avoid reinventing the wheel and use a project that is definitely better
than what I could do.
I had found astronify on github, based on pyo, which seems to be a dead
project, not developed for years, and unfortunately gives me an error
when importing, so I can't use it.
Well, I'm writing to you in the hope that you have read something about
this that you can direct me to. I understand that I am at the edge of
the topic of this list, but... I'll give it a try and cross my fingers.

Thank you in any case and best regards.

--

Gabriele Battaglia (IZ4APU)
--... ...-- -.. . .. --.. ....- .- .--. ..- - ..- . .
Sent from my Giant Desktop PC

[Bibiko]

Hi Gabriele,

oh my goodness, sonification is a really, really broad field. It depends on what you want to hear based on the data available.

If you have Pyo installed on your computer (it can be a little tricky sometimes), you might want to take a look at the following Python script as a trivial and naive starting point, which only works with Pyo. You can "play around" with all the parameters a lot. If you have wxPython installed, you can also do this live, as Pyo comes with two methods: {var_name}.ctrl() and {var_name}.graph(). Read the documentation, but if you don't know Pyo at all, it takes a little time to understand Pyo, but it's worth it :)

-----------------------

from pyo import *
from pyo.lib._wxwidgets import (tFromValue, interpFloat)
import random


s = Server(duplex=1).boot().start()

# any data as a list of floats
data = [random.uniform(1, 5) for x in range(1000)]

# set the lowest and highest output frequency in Hertz
low_freq = 200
high_freq = 800

# set duration and spacing of a note (frequency) in sec
note_duration = .02
note_spacing = .001

# normalize data into a range of 0...1
norm_data = [tFromValue(a, min(data), max(data)) for a in data]

# transform/interpolate normalized data between low_freq and high_freq
freq_data = [interpFloat(a, low_freq, high_freq) for a in norm_data]

# load a pyo table with these frequencies
tab = DataTable(size=len(freq_data), init=freq_data)

# start a clock generator - how quickly the frequencies are played in succession
m = Metro(note_duration+note_spacing).play()

# create a counter to iterate through the data table
c = Counter(m, min=0, max=len(freq_data), dir=0)

# get the frequency from freq_data according to counter
pit = TableIndex(tab, c)

# create an envelope around each played note (frequency) in sync with the clock generator
env = Expseg([
                (0, 0),
                (note_duration/8, 1),
                (note_duration, 0),
                ((note_duration+note_spacing), 0)
            ],
            exp=1., loop=True, mul=.1).play()

# play the current note as a sinus wave in succession
a = Sine(freq=pit, mul=env).mix(2).out()

# if you like add a bit reverb - if so, you can remove '.out()' from Sine(...)
d = WGVerb(a, feedback=[.8, .95], cutoff=high_freq*3, bal=.5, mul=1.).out()

# if you have wxPython installed:
s.gui(locals())

# if you haven't install wxPython:
try:
    while True:
        input('Press ENTER to quit')
        break
except Exception:
    pass

--------------------

Best, Hans

[gabriele....@gmail.com]

Ciao Hans.

Many, many thanks for this fine answer.

Yes I have the basics of Pyo.

I like your script very much and I’ll go to play with it. My idea is to be able to have a parametrized sound. I would like to give the time ax to the stereo field: 0 at the very left and last data at the very right, then pitch linked to the values on y axis.

Astronify allows to to work with pandas and you could pass to it 1 or more columns of your DataFrame. But as far as I know it works in mono by the way and above all it’s an abandoware project.

 

Anyway, again all mygratitude, I’m going to save and study your script.

Gabe.