Grammar Induction for Musical Melodies

Universiy of Southern California, Spring 2007

in ISE575/EE675/CSCI575/PSYCH675

by Reid Swanson (2007)

Code & Documentation

Code

Here are links to the source packages I developed for this project. The only non standard header that I include is unistd.h for command line processing, so it shouldn't be too hard to get to compile (configure, make, make install). The code is written in c++ and compiles successfully on Linux using gcc 4.1.1.

• Extracting: A simple perl script to extract voices from kern data files. It requires the Humdrum toolkit to be installed and you must change the variable 'bin' to point to the installation directory.
• Segmenting: Package used to segment the voices.
• Normalizing: Package used to normalize the segments.
• Parsing: Package used to parse the normalized data.
• Visualization: Package used to convert trees to GraphViz dot files.

Documentation

All documentation is generated by Doxygen. Although there are also a few notes in each package's README file.

• segment API
• normalize API
• parser API
• tree2dot simply takes a tree as input from stdin and outputs a dot file to stdout.

Unrelated Fun Stuff

This is totally unrelated to the project or the class, but that's not going to stop me. While at art school at UCLA (go Bruins, I mean Trojans) I used sound to create a couple of my pieces. I think this piece was particularly successful (at least relatively for my work) and I hope more people can enjoy it than currently do (me). The piece was conveived as a sort of sound sculpture where the sound created a new aural topology to the room it was being played in. The basic idea is that a sinusoidal wave function that varies over time is played out of each speaker of a stereo system. The cancellation pattern will result in a mountainous space of high and low volumes that pulses as you walk in space or as you stand still in time. So take off your headphones and enjoy.

Sound Sculpture

• Part 1
• Part 2
• Part 3