Here’s a neat example of data analysis and visualization (sadly, not using R) with an interesting application: detecting rock songs where the drummer uses a “click track” (an audio metronome that keeps the drum beat to a constant pace). Using a click track makes editing and mixing a song together easier by making it easier to line up beats, but some claim that the robotic drumming style drains the life from a song.
Anyway, Paul Lamere used a Python library to analyze the duration of the beats in mp3 tracks over time (an interesting latent-variable exercise in its own right, but sadly with few details). He then smoothed the beat durations and plotted differences (using gnuplot) from this average over time. Click tracks predate the Beatles, so here’s what Ringo Starr’s beat durations in Dizzy Miss Lizzie look like:
As you can see, the drum beats drift up and down in duration several times over the course of the song. By comparison, the drum beats in a Britney Spears song barely change duration over time:
It’s neat the information you can extract with a bit of data analysis and simple series of charts plotted on the same scale.
Music Machinery: In search of the click track
