NIME 2018

Texton Mirrors
by Erik Nyström

The late neuroscientist Bela Julesz invented the term 'textons' for describing the ‘perceptual atoms’ of texture in visual perception. In the present performance, the idea of textons may be used to describe the spatial distribution of sonic blots, particles, and figures in a 3D sound environment. A montage of textons is created through improvisation and algorithmic processes, where machine learning is used to classify, generate and match sonic morphologies as an extension of performance through artificial intelligence. Mirrored resonances are formed in the multi-dimensional auditory projections of textons, and in the way that machine intelligence creates a posthuman reflection of listening perception.
The texture generating process and performance interface apply neural networks, clustering algorithms, self-organising maps, cellular automata and spatial feedback networks. The work was realised as part of a Leverhulme Fellowship at BEAST (Birmingham Electroacoustic Sound Theatre), University of Birmingham, researching synthesis of spatial texture in composition and performance.