Label reMusik.org is pleased to announce the release of a new work by Sergey Kostyrko — error(msg). This electronic music work was created by use of the experimental ways exploring modular synthesizer, vibromotors, cymbals and drums effects.
Recently, the approaches using artificial neural networks, which are based on the principles of the organization and functioning of biological nervous systems, have started to prevail among the methods of signal processing. The flexibility of adjusting the parameters of such algorithms in the course of their adaptation to the input data makes it possible to solve a variety of problems in spectral analysis, filtering, identification and classification of signals. In general, all the methods of using neural networks, one way or another, could be summarized to prediction, when the algorithm should react in a certain way to changes in the input signal adjusting to its statistical parameters. The indicator of this changes is the error signal, which is formed by comparing the input and output signals. During learning, the neural network refines the parameters of the algorithm in such way as to minimize the resulting error.
The interest of presented work relates to the use of error signals obtained in solving forecasting problems with a linear neural network for constructing a musical composition. As a sound programming environment, a hybrid system “computer-modular synthesizer” is used where such signals allow to control the sound synthesis parameters. To gain the artistic effect, the resulting electrical signal transformed into the mechanical vibrations of different resonators through the use of vibromotors. This approach enhances the element of unpredictability and inaccuracy during sound reproduction.