The Montreux Jazz Festival archive, with its thousands of songs, offers a unique opportunity to discover music. However, the database is so huge that it is difficult to find a way trough this music jungle.
New tools for signal/data analysis and visualization are developed in the lab that could help extract relevant information from this large amount of data.
The project consists in:
- extracting properties from each song spectrogram and creating a set of features characterizing the song,
- creating a graph of songs grouped by similarity,
- infering songs properties such as genre or style from this structure,
- using cutting edge visualization techniques to explore and discover the dataset.
Requirements: A passion for music and some programming skills (python, matlab...).