Based on papers about automatic audio segmentation using a measure of audio novelty
This was made for a final project in a computational audio class in Spring 2017, taught by professor Wayne Snyder at Boston University. Motivation for the methods are in several papers referenced below.
- install Jupyter Notebook (comes with the Anaconda distribution)
- parameters are adjustable at the beginning of Part 1 and may effect run time
- you can try your own music by adding it to the "music/" folder and making the "title" variable the name of the file (must be .wav)
- code is documented and heavily based off of papers referenced below, and can be used as a learning tool
- the .mp4 file is an example rendering of the last code cell in the iPython notebook
- a more detailed description/analysis is in the final report pdf
- cs591utilities.py and parts of Part 1 in novelty_scores.ipynb are provided by Wayne Snyder
- music provided in repository is NOT owned by me, and is purely meant for educational purposes (necessary for running the code)
- http://musicweb.ucsd.edu/~sdubnov/CATbox/Reader/p77-foote.pdf
- https://www.fxpal.com/publications/automatic-audio-segmentation-using-a-measure-of-audio-novelty.pdf
- https://pdfs.semanticscholar.org/85f5/66664bc945f83dca776a322dc337900a8a86.pdf
- https://infoscience.epfl.ch/record/190844/files/2013_Audio%20Novelty-based%20Segmentation%20of%20Music%20Concerts.pdf
- https://quod.lib.umich.edu/cgi/p/pod/dod-idx/music-structural-analysis-via-novelty-shape-detection.pdf?c=icmc;idno=bbp2372.2004.132