Audio Structural Analysis for Music Segmentation

This project aims to explore how state-of-the-art deep learning models can accurately identify and label musical sections like the “verse” and “chorus” across various musical genres. The research will begin by examining available datasets for music structural segmentation, as well as recent deep learning methods suitable for this data. Following the exploration of available data and models, the team will conduct experiments to assess the strengths and capabilities of the different models, proposing enhancements as needed. By training and testing these models, the goal is to simplify music production in digital audio workstations and music notation editors.

Ultimately, this research aims to streamline the creative process for musicians, sound engineers, and composers, making their workflows with audio software faster and smoother by automatically locating and labeling sections of musical compositions.

Faculty Supervisor:

Ichiro Fujinaga

Student:

Partner:

Avid Technologies

Discipline:

Computer science

Sector:

Information and cultural industries; Manufacturing

University:

McGill University

Program: