dc.description.abstract | Harmony is a fundamental structuring principle in Western music, determining
how simultaneously occurring musical notes combine to form chords, and how
successions of chords combine to form chord progressions. Harmony is interesting
to psychologists because it unites many core features of auditory perception
and cognition, such as pitch perception, auditory scene analysis, and statistical
learning. A current challenge is to formalise our psychological understanding
of harmony through computational modelling. Here we detail computational
studies of three core dimensions of harmony: consonance, harmonic expectation,
and voice leading. These studies develop and evaluate computational models
of the psychoacoustic and cognitive processes involved in harmony perception,
and quantitatively model how these processes contribute to music composition.
Through these studies we examine long-standing issues in music psychology,
such as the relative contributions of roughness and harmonicity to consonance
perception, the roles of low-level psychoacoustic and high-level cognitive processes
in harmony perception, and the probabilistic nature of harmonic expectation.
We also develop cognitively informed computational models that are
capable of both analysing existing music and generating new music, with potential
applications in computational creativity, music informatics, and music
psychology. This thesis is accompanied by a collection of open-source software
packages that implement the models developed and evaluated here, which we
hope will support future research into the psychological foundations of musical
harmony.
4 | en_US |