Music Analysis Statistics Quandary
So, for a recent assignment, I've decided to gather data from a variety of songs (similar in Key & BPM) by calculating the sum of each of 12 frequency bands levels through the course of the song. In other words, how much Bass, Mid-tones, and Treble is played independently throughout the track.
I have no issues in regards to the collection of the data, however, my hurdle is finding a meaningful way to compare these values. By calculating a raw sum value for each frequency band of said song, I have the means to compare the songs, but without any type of analytic/modeling process the data has no real substance.
What would be a useful tool/test, analytic procedure, or process of data manipulation that would serve as a legitimate statistical way to compare songs?
Please, any insightful contribution or reasoning would be really cool. It'd be great to see the different approaches to this one.
P.S. Don't hesitate to post highly complex methodologies. I'm really interested in seeing what can be argued as method of comparing something subjective/artistic such as a piece of music.
Re: Music Analysis Statistics Quandary
Are you aware of the field data mining? The techniques and frameworks involved sound right up your alley.
The open-source statistical platform R has a front-end containing a lot of data mining functionality with a GUI interface:
The R Project for Statistical Computing
Togaware: Rattle: A Graphical User Interface for Data Mining using R