Emotional Music Generation Using Interactive Genetic Algorithm

This paper proposes an approach to generate emotional music using Interactive Genetic Algorithm. Performance rules are described by the KTH rule system. Different emotional expressions can be modeled by using selections of rules and rule weights. Interactive genetic algorithm is applied to optimize rule weights based on userspsilasubjective evaluation. Users can get their satisfied emotional music with the help of GA, although they have no profound knowledge for composition. Two kinds of emotional music with the states of happy and sad are obtained finally. The Sheffepsilas method of paired comparisons is carried out to estimate the effectiveness of generated music. In the confidence interval of 99%, the music generated by the proposed approach are more approbatory than that generated by the expert, both for the emotion of happy and sad.