Self-Organizing Topological Timbre Design Methodology Using a Kohonen Neural Network

Generating sounds for music composition with the desired timbral characteristics has been a challenge ever since the dawn of electroacoustic music. Timbre is a remarkably complex phenomenon that has puzzled researchers for a long time. Actually, the nature of musical signals is not fully understood yet. In this paper, we present a sound synthesis technique that uses Kohonen’s one-dimensional self-organizing map to generate neuronal-sounds to respond to a fixed and predefined set of stimulus-sounds, producing timbral variants with the desired characteristics. The self-organizing algorithm provides maintenance of topology so that the intended aesthetical result is properly achieved by avoiding the formal definition of the timbral attributes. To evaluate the obtained results we propose crossing a mathematical/subjective spectral distance from the neuronal-sounds to the stimulus-sounds with the method of timbral classification using Kohonen’s two-dimensional self-organizing map.