A couple weeks ago I watched a show on PBS that discussed whether mathematics existed in the natural world or if we created mathematics to explain the workings of the universe. Music was used as an example.
I’ve always been fascinated with the mathematical relationships in music. So much so I bought a Commodore Amiga computer and Yamaha DX-7 synthesizer and taught myself C Language programming to create apps that would allow the two computers to communicate. ( Yes, that along ago!) In the world of Data Science we use visualizations to imagine what data relationships might look like, I looked at the the data and said why not listen to it ? This series of articles will explore that process through the wonders of MIDI.
This file assigns the values of the climate data to pitches, not even scratching the surface of what can done with effects and more complex note structures. We’ll leave that for a future article.
To begin, I selected a collection of climate data from the Minnesota Dept. of Natural Resources. The particular sample I selected was from 1/1/2010 to 12/31/2010 and contained date, high temp, low temp and precipitation columns. Below is small sample.
2010-01-01,6,-9,T,T,9,24,24,24,0,0 2010-01-02,1,-15,0.00,0.0,9,24,24,24,0,0 2010-01-03,7,-14,0.00,0.0,9,24,24,24,0,0 2010-01-04,7,-10,0.00,0.0,9,24,24,24,0,0 2010-01-05,10,-9,0.00,0.0,9,24,24,24,0,0 2010-01-06,16,-4,0.00,0.0,9,24,24,24,0,0 2010-01-07,16,4,0.06,2.2,11,24,24,24,0,0 2010-01-08,8,-6,0.00,0.0,11,24,24,24,0,0 2010-01-09,6,-14,0.00,0.0,11,24,24,24,0,0 2010-01-10,15,-6,0.00,0.0,11,24,24,24,0,0 2010-01-11,22,8,T,T,10,24,24,24,0,0 2010-01-12,22,2,T,T,10,24,24,24,0,0
Next I coded a Python script that would allow me to read the data rows and build a MIDI music file as the result. I chose Python because of the quick, cross-platform usage. There are several very robust MIDI libraries available for this task written in Java, .Net and other high level languages. The library I chose is MidiUtil-0.89 mostly because it’s written in Python.
First, we had to adjust some of the data values to account for negative values ( cold days ) as MIDI values are 0-128 and increase the values so the pitches where closer to middle C to avoid extremely low pitches that are barely audible. For expediency I accomplished this with inline code but could also be done in R or in a database.
Here is the entire script with comments #
############################################################################
# A sample program read a data file, generate a midi file
# and write to disk.
############################################################################
#Import the library
from midiutil.MidiFile3 import MIDIFile
import csv
# constant values
channel = 0
channel2 = 1
channel3 = 2
track1 = 0
track2 = 1
track3 = 2
time = 0
beats = 540
# indexes to elements of data row
highTemp = 1
lowTemp = 2
precipitation = 3
highTempAdjustment = 20
lowTempAdjustment = 30
# Create the MIDIFile Object with 3 tracks plus names of tracks
MyMIDI = MIDIFile(3)
MyMIDI.addTrackName(track1,time,"Temperature MusicHI")
time = time +1
MyMIDI.addTrackName(track2,time,"Temperature MusicLOW")
time = time +1
MyMIDI.addTrackName(track3,time,"Temperature MusicPrecip")
time = time +1
MyMIDI.addTempo(track1,time, beats)
time = time +1
MyMIDI.addTempo(track2,time, beats)
time = time +1
MyMIDI.addTempo(track3,time, beats)
# set voice (sound) to be played on tracks
# we used General Midi sounds ( see General Midi docs )
time = time +1
MyMIDI.addProgramChange(track1,0, time, 53) # voice 1 = 53
time = time +1
MyMIDI.addProgramChange(track2,1, time, 53) # voice 2 = 53
time = time +1
MyMIDI.addProgramChange(track3,2, time, 119) # cymbal = 119
time = time +1
# open and read each line ( data object ) in file
f = open("climate2010.txt")
for row in csv.reader(f):
# calculate pitch value from temperatures
pitch1 = int(row[highTemp]) + highTempAdjustment
pitch2 = int(row[lowTemp]) + lowTempAdjustment
duration = .5
volume = 100
# add initial tracks
# Add a note. addNote expects the following information:
MyMIDI.addNote(track1,channel,pitch1,time,duration,volume)
time = time +1
MyMIDI.addNote(track2,channel2,pitch2,time,duration,volume)
time = time + 1
if row[precipitation] != "0.00": #got some rain today
pitch3 = 96
MyMIDI.addNote(track3,channel3,pitch3,time,3,90)
#print(row[1])
time = time + 4
# change track 3 to ocean sound for the finale !!
MyMIDI.addProgramChange(track3,2, time, 122) # 122 = Seashore
time = time + 1
MyMIDI.addNote(track3,channel3,40,time,45,100) # let it ring....
# And write it to disk.
binfile = open("climatetemp2010_TempsPrecip.mid", 'wb')
MyMIDI.writeFile(binfile)
binfile.close()
And the resulting file ! Not much to look at – unless you are a real hex-head – just a series of hex values. However with a tool such as Midi Microscope you can view all the events in both hex and carbon-based annotation. A necessity for those in the trade! In an upcoming article I’ll cover some of the tools I use to produce sounds, create MP3 files and edit the raw data.
4D 54 68 64 00 00 00 06 00 01 00 03 03 C0 4D 54 72 6B 00 00 0E 69 00 FF 03 13 54 65 6D 70 65 72 61 74 75 72 65 20 4D 75 73 69 63 48 49 96 40 FF 51 03 01 B2 07 96 40 C0 35 96 40 90 1A 64 83 60 80 1A 64 8B 20 90 15 64 83 60 80 15 64 8B 20 90 1B 64 83 60 80 1B 64 8B 20 90 1B 64 83 60 80 1B 64 8B 20 90 1E 64 83 60 80 1E 64 8B 20 90 24 64 83 60 80 24 64 8B 20 90 24 64 83 60 80 24 64 8B 20 90 1C 64 83 60 80 1C 64 8B 20 90 1A 64 83 60
After this file has been created I played the midi file through a VirtualMidiSynthesizer so the hex data in the file is ‘voiced’ by the settings embedded in the file. I ran the output through another piece of software which captured and saved the output as a MP3 file. Audacity is another open source product I use for this purpose. There is quite a bit of setup required to use all these tools so I’ll cover them in another post.
The following MP3 file is the result played back through a digital software synthesizer. The voices are the high and low temps chasing each other about getting higher and more frenetic in summer ( the middle of cut ). The cymbal sounds for each day there was precipitation. Interesting to hear the lack of rain during summer and autumn. So sit back and enjoy the weather ! If you listen to this using Windows Media Player, turn on the visualization options for an added treat.
The code and data files used in this project are available on GitHub.