Evolutionary Computation Applied to Melody Generation


One of the major goals in the field of computer generated music is capturing the intangible human perception which determines whether or not a melody is "good." Human perception of music is extremely difficult to encode in a computer because doing so requires modeling human experiences which combine to form an individual's perception. Modeling perception is even more difficult when one considers that every person on earth has a unique vantage point. A solution employed by many researchers in the field is selecting one genre of music and building models from "good" selections in that style of music. The principal goal of this research is to determine whether a tree based music model can be used in an evolutionary algorithm fitness function to rate computer generated melodies. Well-known examples of church hymnody were employed in the construction of our music model. The results presented show that tree based music models are effective music critics when combined with application specific genetic operators.


Computer Science

Keywords and Phrases

Computer Generated Sound; Music Generation

Document Type

Article - Journal

Document Version


File Type





© 2004 American Society of Mechanical Engineers (ASME), All rights reserved.

Publication Date

01 Jan 2004

This document is currently not available here.