Please use this identifier to cite or link to this item:
|Title:||Methods to Support End User Design of Arrangement-Level Musical Decision Making|
|Authors:||Martin, Aengus Gabriel|
Interactive Music Systems
|Publisher:||University of Sydney.|
Faculty of Engineering & IT
School of Electrical & Information Engineering
|Abstract:||This thesis is concerned with the study of methods and models to support the design of systems that perform music autonomously, by non-programming end users. Specifically, we address the design of musical agents, which are the central decision making components of such systems and which typically make musical decisions on the time scale of a few seconds. We use the term arrangement-level musical decision making to refer to the activity performed by musical agents. We develop and characterise three separate systems for designing musical agents. The first two are prototypes based on partially observable Markov decision processes and programming by example (PBE), respectively. In each case, we demonstrate the potential of the system but identify significant challenges to making it widely applicable. The third system is called the Agent Designer Toolkit (ADTK) and it is the main contribution of this work. It involves combining PBE with a mechanism whereby a musician can embed musical knowledge into an agent. We show that the ADTK can be used to create agents that convincingly emulate styles of arrangement-level musical decision making in a wide variety of musical contexts, both mainstream and experimental, while requiring only small numbers of examples. The ADTK defines a novel class of constraint-based models of musical decision making. To use these models in performance, a new method was developed, based on binary decision diagrams, for computing musical decisions subject to real-time constraints. The ADTK requires no expertise in conventional computer programming and it can be seamlessly embedded in popular music production software. While we identify certain usability issues with the prototype version, we show the promise of a number of strategies for mitigating them, such as that of providing presets. In addition to its use in music performance, we show the potential of the ADTK for other creative uses such as the generation of new musical ideas.|
|Rights and Permissions:||The author retains copyright of this thesis. It may only be used for the purposes of research and study. It must not be used for any other purposes and may not be transmitted or shared with others without prior permission.|
|Type of Work:||PhD Doctorate|
|Type of Publication:||Doctor of Philosophy Ph.D.|
|Appears in Collections:||Sydney Digital Theses (Open Access)|
This work is protected by Copyright. All rights reserved. Access to this work is provided for the purposes of personal research and study. Except where permitted under the Copyright Act 1968, this work must not be copied or communicated to others without the express permission of the copyright owner. Use the persistent URI in this record to enable others to access this work.
|martin_a_thesis.pdf||Thesis||11.75 MB||Adobe PDF||View/Open|
Items in Sydney eScholarship Repository are protected by copyright, with all rights reserved, unless otherwise indicated.