"Synesthesia is a fancy name for when you experience one of your senses through another". When I was in high school I met a talented muscian who had synesthsia. She would describe how when she played the violin, she saw sparks of light moving in the air along with what she played. Because of this she was able to develop what muscians call 'perfect pitch'-- when you can identify a musical note by name just by listening to it.
The project I wanted to pursue relates to this-- I want to develop something that can take a song, read through its signals and return either a single color or a color palette that corresponds to the song. While this project isnt necessarily something that can outright solve an issue, its a project that more so explores the expression of creativity through machine learning.
The I plan on utilizing 2 datasets: Table of Musical Notes and Their Frequencies and Wavelengths which I converted into a csv, and a csv of HTML color codes, their RBG values and their english names.
Some key features to note are the names, and the RBG values. These can be extremely helpful when determining how similar colors are, and how they may look based on the english description. Image of DF here
import pandas as pd
# reading in our colors.csv file
df_colors = pd.read_csv("colors.csv")
df_colors
| name no space | name | html color code | Red value | Green value | Blue value | |
|---|---|---|---|---|---|---|
| 0 | air_force_blue_raf | Air Force Blue (Raf) | #5d8aa8 | 93 | 138 | 168 |
| 1 | air_force_blue_usaf | Air Force Blue (Usaf) | #00308f | 0 | 48 | 143 |
| 2 | air_superiority_blue | Air Superiority Blue | #72a0c1 | 114 | 160 | 193 |
| 3 | alabama_crimson | Alabama Crimson | #a32638 | 163 | 38 | 56 |
| 4 | alice_blue | Alice Blue | #f0f8ff | 240 | 248 | 255 |
| ... | ... | ... | ... | ... | ... | ... |
| 860 | yellow_orange | Yellow Orange | #ffae42 | 255 | 174 | 66 |
| 861 | yellow_process | Yellow (Process) | #ffef00 | 255 | 239 | 0 |
| 862 | yellow_ryb | Yellow (Ryb) | #fefe33 | 254 | 254 | 51 |
| 863 | zaffre | Zaffre | #0014a8 | 0 | 20 | 168 |
| 864 | zinnwaldite_brown | Zinnwaldite Brown | #2c1608 | 44 | 22 | 8 |
865 rows × 6 columns
The most notable features are the note names, and their frequencies. If there is more information that we need for our csv file like .wav signals, that would be looked into and added onto the csv immediately. Image of DF here
# reading in our note data
df_notes = pd.read_csv("notes_freqs.csv")
df_notes
| Note Name | Octave | Frequency (Hz) | Wavelength (M)* | |
|---|---|---|---|---|
| 0 | C | 0.0 | 16.351 | 20.812 |
| 1 | C# / Db | 0.0 | 17.324 | 19.643 |
| 2 | D | 0.0 | 18.354 | 18.540 |
| 3 | D# / Eb | 0.0 | 19.445 | 17.500 |
| 4 | E | 0.0 | 20.601 | 16.518 |
| ... | ... | ... | ... | ... |
| 124 | G | 9.0 | 12543.856 | 0.027 |
| 125 | G# / Ab | 9.0 | 13289.752 | 0.026 |
| 126 | A | 9.0 | 14080.000 | 0.024 |
| 127 | A# / Bb | 9.0 | 14917.240 | 0.023 |
| 128 | B | 9.0 | 15804.264 | 0.022 |
129 rows × 4 columns
The logistics of this project would involve utilizing scipy.io similarly to how we modified .wav files in lab3. We could assign specific key words with specific chords or notes in songs and that could be cross referenced with the names of colors. Those colors could also include color families, or colors that have a large amount of a specific color value. For example, if we accociate the note Ab with the word 'melancholy', we could then assign 'Melancholy blue' to that note and include colors that have high blue values to that word as well.