Komarov O. Models and methods of dynamic music synthesis based on a genetic approach.

The work is dedicated to the task of automated generation of musical compositions in real time by harmonizing melodies based on a genetic approach. The goal of the dissertation research is to solve the scientific and technical problem of developing models and methods of dynamic musical synthesis with increased harmoniousness of synthesized musical compositions in conditions of limited computing power. The following points of scientific novelty have been achieved: 1. For the first time a consonant chord model of genotypic representation of musical compositions was proposed, the feature of which is compliance with the principle of feasible completeness due to the use of chord level of structural generality, which allowed to reduce the search space in genetic method of harmonization. 2. Improved genetic method of harmonization of musical compositions using models of genotypic representation of musical compositions that are compliant with the principle of feasible completeness, which increased the level of harmoniousness of musical compositions for the selected musical system. 3. For the first time a method of creating models of genotypic representation of musical compositions for arbitrary music system and system with regular mode, the feature of which is compliance with the principle of feasible completeness due to the use of neural autoencoder for achieving structural generality, which allowed to reduce the search space in genetic method of harmonization. 4. For the first time a model for calculating the computational requirements of dynamic music synthesis systems based on a genetic approach was proposed, which allowed to analytically compare the efficiency of different dynamic music synthesis systems and determine required hardware resources to operate within specified time constraints.