Developing Metrics for Evolving a Musically Satisfying Result from a Population of Computer Generated Music Compositions

Creating a computational metric of the quality and musicality of a computer generated music composition is a challenging problem. This paper describes a possible way to develop numerically based metrics for standardizing the fitness evaluation of a population of computer generated music compositions. These metrics are based on assertions by music theorists about the dynamic forces within music that provides the music with a sense of motion and phrasing. The actions of several musical parameters are ranked according to the sense of intensity they produce. This process is applied to an acoustic music composition, and the values are then plotted on a graph. The composite curve from these values give the evaluator a sense of the patterns of lower and higher degrees of intensity, thus providing a sense of how well the music provides a sense of motion and phrasing