Recognition of Musical Dissonance and Consonance in a Simple Neuromorphic Computing System

Reservoir computing with neuromorphic synaptic elements is an emerging, but very successful approach towards processing and classification of various signals. It can be described as a model of a transient computation, where the influence of input changes the internal dynamics of a chosen computational system. Trajectory of these changes represents computation performed by the system. The selection of a suitable computational substrate capable of non-linear response and rich internal dynamics ensures the implementation of simple readout protocols. Signal evolution based on the rich dynamics of the memristive synapse layer helps to emphasize differences between given signals thus enabling their clustering. Here we present a simple neuromorphic computing system (single node echo-state machine based on the memristive synaptic bridge) implemented on the Multisim platform as a tool for clustering of musical intervals according to their consonant or dissonant character. The system generates a series of "epochs" - images of input signal at different stage of evolution. A readout layer based on peak counting in their Fourier spectra allows clustering of musical intervals in a way similar to human subjects or specialized algorithms. The result of this data evolution closely resembled the sensory dissonance curve, with some significant differences. Interestingly, clustering is performed without any reference to the theory of music. This study shows a high potential for exploiting a simple neuromorphic system for advanced information processing. Furthermore, they indicate that the notions of consonance and dissonance may have the neurophysiological background.