Skip to content

Latest commit

 

History

History
29 lines (20 loc) · 1.75 KB

README.md

File metadata and controls

29 lines (20 loc) · 1.75 KB

Music Similarity

The goal of this project is to research how similar Shōnen/ Shōjo Anime openings are.
The system uses song signal level features to check the similarity of songs. In short, a song is modelled as a Gaussian Mixture Model (GMM) where the "similarity" of two songs is measured as the "distance" between the GMMs. Multiple types of distance measures can be used, but most examples here use the Jenson-Shannon Divergence.

A custom dataset of 238 songs from the reddit AnimeThemes wiki was used for training where the songs were categorized using MyAnimeList data.

Demo

A full example of the project is given in this notebook.
It includes an example on training and using the music similarity model.

Installation

Note that this is a toy project and should not actually be used for anything production related.

The package can be installed by cloning the repository and installing it with pip.

$ git clone https://github.com/nickyu42/music-similarity
$ cd music-similarity
$ pip install .

References

  • Mandel, M. I., & Ellis, D. P. (2005). Song-level features and support vector machines for music classification.
  • Schnitzer, D., Flexer, A., Schedl, M., & Widmer, G. (2011). Using Mutual Proximity to Improve Content-Based Audio Similarity. In ISMIR (Vol. 11, pp. 79-84).
  • Jensen, J. H., Ellis, D. P., Christensen, M. G., & Jensen, S. H. (2007). Evaluation distance measures between gaussian mixture models of mfccs.
  • Tzanetakis, G. (n.d.). Tutorial: Music Information Retrieval. http://www.marsyas.info/icmr2014_tutorial_2_per_page.pdf