Section: New Results

Music Content Processing and Information Retrieval

Music structure, music language modeling, System & Contrast model

Current work developed in our research group in the domain of music content processing and information retrieval explore various information-theoretic frameworks for music structure analysis and description [24], in particular the System & Contrast model [1].

Tensor-based Representation of Sectional Units in Music

Participants : Corentin Guichaoua, Frédéric Bimbot.

Following Kolmogorov'€™s complexity paradigm, modeling the structure of a musical segment can be addressed by searching for the compression program that describes as economically as possible the musical content of that segment, within a given family of compression schemes.

In this general framework, packing the musical data in a tensor-derived representation enables to decompose the structure into two components : (i) the shape of the tensor which characterizes the way in which the musical elements are arranged in an n-dimensional space and (ii) the values within the tensor which reflect the content of the musical segment and minimize the complexity of the relations between its elements.

This approach is currently developed and tested for the grouping of chord sequences into sectional units for pop music songs, with very encouraging segmentation results on pop songs.

Minimal Transport Graphs for the Modeling of Chord Progressions

Participants : Corentin Louboutin, Frédéric Bimbot.

In this work, we model relations between chords by minimal transport and we investigate different types of dependencies within chord sequences [33]. For this purpose we use the €œSystem & Contrast (S&C) model [1], designed for the description of music sectional units, to infer non-sequential structures called chord progression graphs (CPG).

Minimal transport is defined as the shortest displacement of notes, in semitones, between a pair of chords. The paper [33] present three algorithms to find CPGs for chords sequences: one is sequential, and two others are based on the S& C model. The three methods are compared using the perplexity as an efficiency measure.

The experiments on a corpus of 45 segments taken from songs of multiple genres indicate that optimization processes based on the S&C model outperform the sequential model with a decrease in perplexity over 1.0.

Regularity Constraints for the Fusion of Music Structure Segmentation System

Participant : Frédéric Bimbot.

Main collaborations Gabriel Sargent (EPI LinkMedia, Rennes, France)

Music structure estimation has recently emerged as a central topic within the field of Music Information Retrieval. Indeed, as music is a highly structured information stream, knowledge of how a music piece is organized represents a key challenge to enhance the management and exploitation of large music collections.

Former work carried out in our group [9] has illustrated the benefits that can be expected from a regularity constraint on the structural segmentation of popular music pieces : a constraint which favors structural segments of comparable size provides a better conditioning of the boundary estimation process.

As a further investigation, we have explored the benefits of the regularity constraint as an efficient way for combining the outputs of a selection of systems presented at MIREX between 2010 and 2015. These experiments have yielded a level of performance which is competitive to that of the state-of-the-art on the ”MIREX10" dataset (100 J-Pop songs from the RWC database) [18].