Melody Harmonization in Evolutionary Music Using Multiobjective Genetic Algorithms

This paper describes a multiobjective approach for melody harmonization in evolutionary music. There are numerous methods and a myriad of results to a process of harmonization of a given melody. Some implicit rules can be extracted from musical theory, but some harmonic aspects can only be defined by preferences of a composer. Thus, a multiobjective approach may be useful to allow an evolutionary process to find a set of solutions that represent a trade-off between the rules in different objective functions. In this paper, a multiobjective evolutionary algorithm defines chord changes with differing degrees of simplicity and dissonance. While presenting such an algorithm, we discuss how to embed musical cognizance in Genetic Algorithms in a meta-level. Experiments were held and compared to human judgment of the results. The findings suggest that it is possible to devise a fitness function which reflects human intentions for harmonies.