Projet de recherche doctoral numero :7347


Date depot: 19 juin 2020
Titre: Computational Detection & Shaping of Emergent Musical Structures
Directrice de thèse: Elaine CHEW (King's college London)
Directeur de thèse: Carlos AGON (STMS)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini

Resumé: This research addresses problem solving skills in the perception and cognition of music structures in performed music, taking advantage of crowdsourcing to amass large amounts of data that require human knowledge and intelligence for analysis. A secondary goal will be to investigate decision making in the act of creating performed structures, developing bespoke software-based sandbox environments to experiment with making and shaping performed structures. The research brings together music performance studies, music perception and cognition, and music information research to develop a novel methodological framework for computational music structure analysis with primary focus on structures experienced and made in music performance that will inform current efforts in music and artificial intelligence. The mathematical tools will be based on and combine methods in machine learning, optimization, topology and statistics. In fact, the detection of musical patterns from performed music data as well as score information allows for a better understanding of the musical structure. Mathematical morphology has been used successfully in the recognition of shapes in images. We will extend this family of techniques and adapted them to music features such as melodic shapes and performed articulations to automatically detect figural groupings. Moreover, a range of representations of music performance and score information will be designed and studied using mathematical theories such as persistent homology from algebraic topology, which allows for the detection of persistent features over different spatial and time scales. One of the longer-term objectives concerns the irregular rhythms in electrocardiographic (ECG) traces of arrhythmias. In fact, they also exhibit grouping structures that lend themselves readily to the kinds of analyses designed for automated analysis of performed music. Thus, the techniques that will be used and developed for performed music will also be applied to ECG recordings of arrhythmias to help characterise patterns that can serve as biomarkers for disease progression or subtypes.

Doctorant.e: Bedoya Ramos Daniel Augusto