Synthesis and perception of musical timbre

Timbre is an essential component in music that describes all tonal qualities and characteristics of sounds, excluding their fundamental pitch and loudness. However, timbre is a vaguely understood property of sound due to its perceptual complexity and the presence of imprecise frameworks that superficially communicate timbral concepts across different cultures, disciplines, and individuals. Many mechanisms of human sound perception, culture, and individual preferences influence our holistic perception of timbre, making it challenging to measure and explain. Yet, timbre is one of music's most important properties that musicians and music producers can use to express feeling and creativity. Understanding timbre and its many facets is crucial for mixing different instruments and sounds or designing individual sounds in production or live performance.
Fritz et al. [1] demonstrated that the timbre of a violin could be described using descriptors such as "bright" or "warm"; furthermore, they successfully correlated these descriptors with specific adjustments to the frequency spectrum of sound that violins produce. With inspiration from Fritz et al., this research aims to develop a taxonomy of semantic timbral descriptors used in orchestral performance and music production tools (e.g. equalisation (EQ) and compression plugins) to subsequently develop new culturally enhanced artificial intelligence (AI) models of timbre perception, similarity, and transfer that have so far mainly focused on timbre as an acoustic phenomenon and only consider temporal or spectral characteristics. These culturally informed and enhanced timbre models shall then be applied towards designing, developing, and evaluating a human-centred software package that can classify and generate different examples of timbre based on the relevant timbre descriptors. The research phases shall heed the steps of the user-centred design process, and user research methods such as semi-structured interviewing and...