Semantic Design Methodologies for Acoustic Modelling Interfaces

While there have been developments in semantic interfaces for digital audio effects in order to allow non experts in audio engineering to understand and apply audio transforms within an audio production workflow, there is relatively less research available in applying these principles towards interfaces that allow non experts to treat live environments to induce desired perceptual factors of sound played within them. The aim of this research was to propose methods to derive associations between semantic descriptors of audio that would be used within a hypothetical semantic interface, and transformations of an acoustic environment. These methods were then investigated for their effectiveness via the deriving associations for two selected perceptual factors, 'brightness' and 'closeness', in relation to two room modification variables, 'source/receiver distance' and 'absorption coefficient of wall surfaces'. The hypothetical interface was conceptualised as a three schema arrangement; with the perceptual terms in the external layer, acoustic measurement in the conceptual layer, and room modification variables in the internal layer. Experiments with auralisation models were performed to derive associations between perceptual terms and acoustic measurements, while a listening test was undertaken to derive associations between acoustic measurements and room modification variables. The results from these two experiments were cross referenced with each other to derive associations between perceptual terms and room modification variables. It was concluded that absorption coefficient was the primary factor influencing perceptual brightness, while source/receiver distance was the primary factor influencing perceptual closeness. It is recommended that these methods be further investigated for more complex associations to be made for a greater variety of perceptual terms to move towards the development of a semantic interface that can be used with acoustic models. Future applications of these methodologies could integrate machine learning methods to allow for more complex systems and a greater range of applications.