##vox

I would like to create a visualization of data for voices, that breaks them down into informative components. There are a few steps involved here:

• Getting a bunch of songs
• Isolating (to some extent) the vocals on those songs/podcasts are easy
  • this
  • and this
• Extracting features from the voices
  • Something like yaffe
• Doing some dimensionality reduction/interpretation to get at what the features mean intuitively
  • Use extremely unique voices for this purpose (Q-Tip, Meechy Darko, Roman Mars, Paul Seufert)
• Presenting the data nicely
  • Polygon charts (like the ones used for multiple intelligence tests)