Abstract
The purpose of this paper is to implement a system to compose melodies that reflect user's feeling using the Interactive Genetic Algorithm (IGA). One melody corresponds to a chromosome containing information on tone color, tune, the number of notes and the length of notes. Melodies are composed by the GA operations of 200 chromosomes. Twenty melodies are chosen by objective fitness functions that consider musical basic rules. A user listens to the chosen melodies and evaluates them. The chromosome with the best evaluation becomes an elite. The strategy of elitist selection and that of roulette are used together in GA. The procedures are repeated until a user finds a melody that fits with the user's feeling. Experimental results show that the procedures with the fitness functions reduce human fatigue caused by the incorporation of human evaluation into GA and that the presented system composes melodies reflecting user's feeling well.
