A while back I wrote about a crazy scheme to generate perfect music. I thought about it for a while but never really got anywhere with it. Time to start the “research” phase, if you could call it that. I won’t make it past this step since I don’t have a significant background in music theory, so I won’t be able to analyze the more important musical aspects of these songs.
So here’s a summary of the idea. Music is just a series of different tones. When played in succession and continuously, you’ve got yourself a trippy techno beat, sappy love song or energizing march. So the idea, inspired by the perfect face algorithm, is that universally (95%-99%) pleasing music must have some sort of similarity that makes them pleasing to such a majority of the population.
For now the experiment will be limited to just one culture – American. Obviously we can’t throw in some chart topping Bollywood beats or something. So the description universally appealing will indicate a song that is a chart topper on the Billboard Hot 100. Here’s some more on that. The chart topping songs are selected by sales and airtime. Initially I thought this would be a problem, as only a particular type of person would call up a radio or buy a CD, but it’s probably not a significant demographic group that it would interfere with the results.
Here are some selected some songs that have reached the #1 spot on the Billboard Hot 100 to compare.
- Flo Rida – Right Round
- Crank That (Soulja Boy) – Soulja Boy Tell ‘Em
- Whatever You Like – T.I.
- Viva La Vida – Coldplay
- Just Dance – Lady GaGa Ft. Colby O’Donis
- Low – Flo Rida
There’s a lot of hip-hop on the Hot 100, but I drew extra songs from other genres. Here’s the game plan, or the aspects of the songs to be compared.
- Song Name
- Number of weeks on chart
- Genre
- Theme (Ex. What is the song mostly about?)
- Synopsis (Ex. Full blown summary of the lyrics)
- Tempo
- My opinion of the song
- Syncopation
- Chorus (Ex. Yes or no?)
- Time Signature
- Structure (Ex. ABCDABAB)
- Artist (Expand as needed) – Gender, vocal range, etc.
- Instrument – Augmented, etc.
I had more features to analyze before. Can’t think of any more at the moment.
Very interesting. As a shrink, I did muscle relaxation training with many people and have worked, on and off, with finding the “perfect noise” (as opposed to music,) to augment or drive the voluntary relaxation process. That’s much more limited than what you’re writing about, but I notice several factors in common with your list. Being an old fart, I have difficulty imagining myself liking anything which includes hip-hop or rap, but that’s why one does experiments and develops algorithms, isn’t it? There might very well be an element or two that I’m not considering or which I am pre-judging that could change my mind or appeal to me without my ever realizing what it was. I’ll be interested in following this if you pursue it.
Peace, Doc
To actually continue pursuing this project, I would need to learn music theory as well as computer science related topics.