Title: A machine learning approach to computer modeling of musical expression for performance learning and practice
Supervisor: Dr. Rafael Ramírez
Jury: Dr. José Manuel Iñesta (Universidad de Alicante), Dr. Maarten Grachten (Independent consultant), Dr. Renee Timmers (University of Sheffield)
This thesis deals with the design and implementation of computer systems for expressive music performance (CSEMP), exploring different methods from machine learning and reflecting on the role of musical structure in the emergence of performance patterns, as well as the applicability of each approach in a pedagogical setting. Three models are described and evaluated: a lazy learning approach using a phrase similarity measure, an evolution of the previous with parameterized performance features, and a deep-learning model with sequential encoding of musical information. Results demonstrate that the simpler phrase-level approaches can generate stimulating performances with small datasets, and that the deep-learning approach can achieve high accuracy predicting performance information. Their analyses also highlight the challenges of designing systems for instruments beyond the piano. The pedagogical potential of technologically-enhanced settings is addressed with the proposal and pilot evaluation of a performance practice method using the SkyNote software.
This thesis defense will take place online. To attend use this link (ID of the meeting 813 3124 4226). The microphone and camera must be turned off, and the online access will be unavailable after 30 minutes from the start of the defense.