Interfacing Sounds: Hierarchical Audio-Content Morphologies for Creative Re-purposing in earGram

Abstract

Audio content-based processing has become a pervasive methodology for techno-fluent musicians. System architectures typically create thumbnail audio descriptions, based on signal processing methods, to visualize, retrieve and transform musical audio efficiently. Towards enhanced us-ability of these descriptor-based frameworks for the music community, the paper advances a minimal content-based audio description scheme, rooted on primary musical notation attributes at the threefold sound object, meso and macro hierarchies. Multiple perceptually-guided viewpoints from rhythmic, harmonic, timbral and dynamic attributes define a discrete and finite alphabet with minimal formal and subjective assumptions using unsupervised and user-guided methods. The Factor Oracle automaton is then adopted to model and visualize temporal morphology. The generative musical applications enabled by the descriptor-based framework at multiple structural hierarchies are discussed.